diff --git a/.gitattributes b/.gitattributes index 27499b90cdaefa2ef1434b4e780a0257c61ef0c7..d29beb4a23197277ea146e7ac1eacde5a1b5edc3 100644 --- a/.gitattributes +++ b/.gitattributes @@ -45,3 +45,4 @@ qwen35-08b-fp16-int4-g32-gptq-home/graph.json filter=lfs diff=lfs merge=lfs -tex qwen35-08b-fp16-int4-g32-gptq-home/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text qwen35-08b-fp16-int4-g8-gptq-up-proj-exclude-l15-candidate/graph.json filter=lfs diff=lfs merge=lfs -text qwen35-08b-fp16-int4-g8-gptq-up-proj-exclude-l15-candidate/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text +qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/graph.json filter=lfs diff=lfs merge=lfs -text diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/entrypoints/prefill-4.json b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/entrypoints/prefill-4.json new file mode 100644 index 0000000000000000000000000000000000000000..d23e48dbdd4274fd876add04fdcec93942346ae0 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/entrypoints/prefill-4.json @@ -0,0 +1,99057 @@ +{ + "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, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 1017118720, + "parts": [ + { + "path": "weights/00000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00001.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00003.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00004.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00005.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00006.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00007.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00008.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00009.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00010.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00011.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00012.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00013.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00014.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00015.bin", + "offset": 0, + "byteLength": 10485760 + } + ] + }, + "p_model_model_language_model_layers_0_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 10485760, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 10489856, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 10489888, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 10489920, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 35655744, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 44044352, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 44109888, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 44175424, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 44224576, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 44224832, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 52613440, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00015.bin", + "offset": 52617536, + "byteLength": 14491328 + }, + { + "path": "weights/00016.bin", + "offset": 0, + "byteLength": 10674496 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00016.bin", + "offset": 10674496, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00016.bin", + "offset": 16965952, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_1_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00016.bin", + "offset": 42131776, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00016.bin", + "offset": 42135872, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00016.bin", + "offset": 42135904, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00016.bin", + "offset": 42135936, + "byteLength": 24972928 + }, + { + "path": "weights/00017.bin", + "offset": 0, + "byteLength": 192896 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 192896, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 8581504, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 8647040, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 8712576, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 8761728, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 8761984, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 17150592, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 17154688, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 42320512, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00017.bin", + "offset": 48611968, + "byteLength": 18496896 + }, + { + "path": "weights/00018.bin", + "offset": 0, + "byteLength": 6668928 + } + ] + }, + "p_model_model_language_model_layers_2_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 6668928, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 6673024, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 6673056, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 6673088, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 31838912, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 40227520, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 40293056, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 40358592, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 40407744, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 40408000, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 48796608, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00018.bin", + "offset": 48800704, + "byteLength": 18308160 + }, + { + "path": "weights/00019.bin", + "offset": 0, + "byteLength": 6857664 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 6857664, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 13149120, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_3_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 38314944, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": { + "shape": [ + 4096, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 16777216, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 38319040, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 55096256, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 55096768, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 57193920, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 57194432, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00019.bin", + "offset": 59291584, + "byteLength": 7817280 + }, + { + "path": "weights/00020.bin", + "offset": 0, + "byteLength": 571328 + } + ] + }, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 571328, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 575424, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 25741248, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 32032704, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_4_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 57198528, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 57202624, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 57202656, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00020.bin", + "offset": 57202688, + "byteLength": 9906176 + }, + { + "path": "weights/00021.bin", + "offset": 0, + "byteLength": 15259648 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 15259648, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 23648256, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 23713792, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 23779328, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 23828480, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 23828736, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 32217344, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 32221440, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 57387264, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00021.bin", + "offset": 63678720, + "byteLength": 3430144 + }, + { + "path": "weights/00022.bin", + "offset": 0, + "byteLength": 21735680 + } + ] + }, + "p_model_model_language_model_layers_5_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 21735680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 21739776, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 21739808, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 21739840, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 46905664, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 55294272, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 55359808, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 55425344, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 55474496, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 55474752, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 63863360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00022.bin", + "offset": 63867456, + "byteLength": 3241408 + }, + { + "path": "weights/00023.bin", + "offset": 0, + "byteLength": 21924416 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00023.bin", + "offset": 21924416, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00023.bin", + "offset": 28215872, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_6_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00023.bin", + "offset": 53381696, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00023.bin", + "offset": 53385792, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00023.bin", + "offset": 53385824, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00023.bin", + "offset": 53385856, + "byteLength": 13723008 + }, + { + "path": "weights/00024.bin", + "offset": 0, + "byteLength": 11442816 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 11442816, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 19831424, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 19896960, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 19962496, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 20011648, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 20011904, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 28400512, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 28404608, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 53570432, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00024.bin", + "offset": 59861888, + "byteLength": 7246976 + }, + { + "path": "weights/00025.bin", + "offset": 0, + "byteLength": 17918848 + } + ] + }, + "p_model_model_language_model_layers_7_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 17918848, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": { + "shape": [ + 4096, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 16777216, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 17922944, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 34700160, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 34700672, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 36797824, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 36798336, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 38895488, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 47284096, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00025.bin", + "offset": 47288192, + "byteLength": 19820672 + }, + { + "path": "weights/00026.bin", + "offset": 0, + "byteLength": 5345152 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 5345152, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 11636608, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_8_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 36802432, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 36806528, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 36806560, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 36806592, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00026.bin", + "offset": 61972416, + "byteLength": 5136448 + }, + { + "path": "weights/00027.bin", + "offset": 0, + "byteLength": 3252160 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 3252160, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 3317696, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 3383232, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 3432384, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 3432640, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 11821248, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 11825344, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 36991168, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00027.bin", + "offset": 43282624, + "byteLength": 23826240 + }, + { + "path": "weights/00028.bin", + "offset": 0, + "byteLength": 1339584 + } + ] + }, + "p_model_model_language_model_layers_9_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 1339584, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 1343680, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 1343712, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 1343744, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 26509568, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 34898176, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 34963712, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 35029248, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 35078400, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 35078656, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 43467264, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00028.bin", + "offset": 43471360, + "byteLength": 23637504 + }, + { + "path": "weights/00029.bin", + "offset": 0, + "byteLength": 1528320 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 1528320, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 7819776, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_10_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 32985600, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 32989696, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 32989728, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 32989760, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 58155584, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 66544192, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 66609728, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 66675264, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 66724416, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00029.bin", + "offset": 66724672, + "byteLength": 384192 + }, + { + "path": "weights/00030.bin", + "offset": 0, + "byteLength": 8004416 + } + ] + }, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00030.bin", + "offset": 8004416, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00030.bin", + "offset": 8008512, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00030.bin", + "offset": 33174336, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00030.bin", + "offset": 39465792, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_11_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00030.bin", + "offset": 64631616, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": { + "shape": [ + 4096, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 16777216, + "parts": [ + { + "path": "weights/00030.bin", + "offset": 64635712, + "byteLength": 2473152 + }, + { + "path": "weights/00031.bin", + "offset": 0, + "byteLength": 14304064 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 14304064, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 14304576, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 16401728, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 16402240, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 18499392, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 26888000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 26892096, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 52057920, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00031.bin", + "offset": 58349376, + "byteLength": 8759488 + }, + { + "path": "weights/00032.bin", + "offset": 0, + "byteLength": 16406336 + } + ] + }, + "p_model_model_language_model_layers_12_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 16406336, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 16410432, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 16410464, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 16410496, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 41576320, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 49964928, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 50030464, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 50096000, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 50145152, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 50145408, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 58534016, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00032.bin", + "offset": 58538112, + "byteLength": 8570752 + }, + { + "path": "weights/00033.bin", + "offset": 0, + "byteLength": 16595072 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00033.bin", + "offset": 16595072, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00033.bin", + "offset": 22886528, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_13_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00033.bin", + "offset": 48052352, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00033.bin", + "offset": 48056448, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00033.bin", + "offset": 48056480, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00033.bin", + "offset": 48056512, + "byteLength": 19052352 + }, + { + "path": "weights/00034.bin", + "offset": 0, + "byteLength": 6113472 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 6113472, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 14502080, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 14567616, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 14633152, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 14682304, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 14682560, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 23071168, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 23075264, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 48241088, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00034.bin", + "offset": 54532544, + "byteLength": 12576320 + }, + { + "path": "weights/00035.bin", + "offset": 0, + "byteLength": 12589504 + } + ] + }, + "p_model_model_language_model_layers_14_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 12589504, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 12593600, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 12593632, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 12593664, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 37759488, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 46148096, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 46213632, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 46279168, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 46328320, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 46328576, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 54717184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00035.bin", + "offset": 54721280, + "byteLength": 12387584 + }, + { + "path": "weights/00036.bin", + "offset": 0, + "byteLength": 12778240 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 12778240, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 19069696, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_15_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 44235520, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": { + "shape": [ + 4096, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 16777216, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 44239616, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 61016832, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 61017344, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 63114496, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 63115008, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00036.bin", + "offset": 65212160, + "byteLength": 1896704 + }, + { + "path": "weights/00037.bin", + "offset": 0, + "byteLength": 6491904 + } + ] + }, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 6491904, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 6496000, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 31661824, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 37953280, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_16_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 63119104, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 63123200, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 63123232, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00037.bin", + "offset": 63123264, + "byteLength": 3985600 + }, + { + "path": "weights/00038.bin", + "offset": 0, + "byteLength": 21180224 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 21180224, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 29568832, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 29634368, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 29699904, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 29749056, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 29749312, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 38137920, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 38142016, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00038.bin", + "offset": 63307840, + "byteLength": 3801024 + }, + { + "path": "weights/00039.bin", + "offset": 0, + "byteLength": 2490432 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 2490432, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_17_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 27656256, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 27660352, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 27660384, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 27660416, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 52826240, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 61214848, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 61280384, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 61345920, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 61395072, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00039.bin", + "offset": 61395328, + "byteLength": 5713536 + }, + { + "path": "weights/00040.bin", + "offset": 0, + "byteLength": 2675072 + } + ] + }, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 2675072, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 2679168, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 27844992, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 34136448, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_18_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 59302272, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 59306368, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 59306400, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00040.bin", + "offset": 59306432, + "byteLength": 7802432 + }, + { + "path": "weights/00041.bin", + "offset": 0, + "byteLength": 17363392 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 17363392, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 25752000, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 25817536, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 25883072, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 25932224, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 25932480, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 34321088, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 34325184, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 59491008, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00041.bin", + "offset": 65782464, + "byteLength": 1326400 + }, + { + "path": "weights/00042.bin", + "offset": 0, + "byteLength": 23839424 + } + ] + }, + "p_model_model_language_model_layers_19_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 23839424, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": { + "shape": [ + 4096, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 16777216, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 23843520, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 40620736, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 40621248, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 42718400, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 42718912, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 44816064, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 53204672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00042.bin", + "offset": 53208768, + "byteLength": 13900096 + }, + { + "path": "weights/00043.bin", + "offset": 0, + "byteLength": 11265728 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00043.bin", + "offset": 11265728, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00043.bin", + "offset": 17557184, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_20_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00043.bin", + "offset": 42723008, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00043.bin", + "offset": 42727104, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00043.bin", + "offset": 42727136, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00043.bin", + "offset": 42727168, + "byteLength": 24381696 + }, + { + "path": "weights/00044.bin", + "offset": 0, + "byteLength": 784128 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 784128, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 9172736, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 9238272, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 9303808, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 9352960, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 9353216, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 17741824, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 17745920, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 42911744, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00044.bin", + "offset": 49203200, + "byteLength": 17905664 + }, + { + "path": "weights/00045.bin", + "offset": 0, + "byteLength": 7260160 + } + ] + }, + "p_model_model_language_model_layers_21_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 7260160, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 7264256, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 7264288, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 7264320, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 32430144, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 40818752, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 40884288, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 40949824, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 40998976, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 40999232, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 49387840, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00045.bin", + "offset": 49391936, + "byteLength": 17716928 + }, + { + "path": "weights/00046.bin", + "offset": 0, + "byteLength": 7448896 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 7448896, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 13740352, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_22_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 38906176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 38910272, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 32, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 38910304, + "byteLength": 32 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 38910336, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00046.bin", + "offset": 64076160, + "byteLength": 3032704 + }, + { + "path": "weights/00047.bin", + "offset": 0, + "byteLength": 5355904 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 5355904, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 5421440, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 49152, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 5486976, + "byteLength": 49152 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 256, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 5536128, + "byteLength": 256 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 5536384, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 13924992, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 13929088, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 39094912, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00047.bin", + "offset": 45386368, + "byteLength": 21722496 + }, + { + "path": "weights/00048.bin", + "offset": 0, + "byteLength": 3443328 + } + ] + }, + "p_model_model_language_model_layers_23_input_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 3443328, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": { + "shape": [ + 4096, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 16777216, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 3447424, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 20224640, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 20225152, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 22322304, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": { + "shape": [ + 512, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 2097152, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 22322816, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": { + "shape": [ + 2048, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 8388608, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 24419968, + "byteLength": 8388608 + } + ] + }, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 32808576, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 32812672, + "byteLength": 25165824 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": { + "shape": [ + 6144, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 8, + "scales": "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 57978496, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": { + "shape": [ + 2048, + 6144 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 25165824, + "parts": [ + { + "path": "weights/00048.bin", + "offset": 64269952, + "byteLength": 2838912 + }, + { + "path": "weights/00049.bin", + "offset": 0, + "byteLength": 22326912 + } + ] + }, + "p_model_model_language_model_norm_weight": { + "shape": [ + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 22326912, + "byteLength": 4096 + } + ] + }, + "b_model_model_language_model_rotary_emb_inv_freq": { + "shape": [ + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 128, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 22331008, + "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": "float16", + "role": "input", + "byteLength": 32768 + }, + "states_0": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_1": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_2": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_3": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_4": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_5": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_8": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_9": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_11": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_12": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_13": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_14": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_15": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_16": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_17": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_19": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_20": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_21": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_22": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_23": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_24": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_25": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_27": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_28": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_29": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_30": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_31": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_32": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_33": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_34": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_35": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_36": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_37": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_38": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_39": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_40": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_41": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_42": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_43": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "input", + "byteLength": 49152 + }, + "states_45": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_46": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_47": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "input", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "embedding": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "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 + }, + "_to_copy_1": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "_to_copy_2": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "_to_copy_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_1": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_4": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_1": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_13": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_1": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_2": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_3": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_1": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_5": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_11": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_12": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_14": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_14": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_1": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_2": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_3": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_6": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_1": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_3": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_1": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_6": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_7": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_2": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_4": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_2": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_8": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_7": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_8": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_9": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_10": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_11": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_4": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_5": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_12": { + "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 + }, + "_to_copy_13": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_10": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_14": { + "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 + }, + "_to_copy_15": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_4": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_16": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_17": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_8": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_1": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_5": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_2": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_6": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_14": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_7": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_18": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_4": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_19": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_11": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_16": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_2": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_8": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_16": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_7": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_10": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_11": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_2": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_1": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_6": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_1": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_19": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_3": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_20": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_17": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_21": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_22": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_8": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_9": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_10": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_1": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_20": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_21": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_3": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_13": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_6": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_19": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_20": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_4": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_14": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_7": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_21": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_22": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_23": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_24": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_25": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_26": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_11": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_12": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_27": { + "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 + }, + "_to_copy_28": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_23": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_29": { + "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 + }, + "_to_copy_30": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_12": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_16": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_31": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_32": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_18": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_26": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_13": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_5": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_14": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_27": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_19": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_33": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_7": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_34": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_21": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_29": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_4": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_16": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_19": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_17": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_14": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_18": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_19": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_3": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_2": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_7": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_2": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_27": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_6": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_28": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_20": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_29": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_30": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_31": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_15": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_16": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_17": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_2": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_35": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_36": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_5": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_23": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_11": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_32": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_33": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_6": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_24": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_12": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_34": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_37": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_38": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_39": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_40": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_41": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_18": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_19": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_42": { + "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 + }, + "_to_copy_43": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_36": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_44": { + "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 + }, + "_to_copy_45": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_20": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_20": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_26": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_46": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_47": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_28": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_39": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_5": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_21": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_8": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_22": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_40": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_29": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_48": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_10": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_49": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_31": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_42": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_24": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "view_21": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "slice_34": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_35": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_3": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_unsafe_view": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_22": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_50": { + "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 + }, + "_to_copy_51": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_22": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_25": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_23": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "_to_copy_52": { + "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 + }, + "_to_copy_53": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_23": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "linear_26": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_24": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_24": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_8": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "unsqueeze_9": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_36": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_37": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 12288 + }, + "slice_38": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_39": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 3072 + }, + "mul_47": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_40": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "slice_41": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "neg_3": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "cat_4": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_48": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "add_36": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_49": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_42": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_43": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "neg_4": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "cat_5": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "mul_50": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "add_37": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "cat_6": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "cat_7": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "cache_update": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_1": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_10": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_2": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_1": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "unsqueeze_11": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_3": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_2": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "transpose_25": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "matmul_1": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "mul_51": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "add_38": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "softmax": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_to_copy_54": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "matmul_2": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_26": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_6": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_25": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_52": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_27": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_39": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_55": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_56": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_41": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_54": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_28": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_9": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_29": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_55": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_30": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_42": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_57": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_14": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_58": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_44": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_57": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_10": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_31": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_27": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_32": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_26": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_33": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_34": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_8": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_3": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_12": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_3": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_45": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_10": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_46": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_28": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_47": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_48": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_49": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_27": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_28": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_29": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_4": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_59": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_3": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_60": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_7": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_46": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_20": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_60": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_61": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_8": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_47": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_21": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_62": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_61": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_62": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_63": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_64": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_65": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_30": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_31": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_66": { + "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 + }, + "_to_copy_67": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_64": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_68": { + "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 + }, + "_to_copy_69": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_32": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_35": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_49": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_70": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_16": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_71": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_51": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_67": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_11": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_36": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_12": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_37": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_68": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_38": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_52": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_72": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_17": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_73": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_54": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_70": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_12": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_39": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_30": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_40": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_33": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_41": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_42": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_9": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_52": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_4": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_13": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_4": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_53": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_13": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_54": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_31": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_55": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_56": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_57": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_34": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_35": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_36": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_5": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_74": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_4": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_75": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_9": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_56": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_25": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_73": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_74": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_10": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_57": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_26": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_75": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_76": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_77": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_78": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_79": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_80": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_37": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_38": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_81": { + "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 + }, + "_to_copy_82": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_77": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_83": { + "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 + }, + "_to_copy_84": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_39": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_43": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_59": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_85": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_19": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_86": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_61": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_80": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_44": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_15": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_45": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_81": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_46": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_62": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_87": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_20": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_88": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_64": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_83": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_14": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_47": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_33": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_48": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_40": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_49": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_50": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_10": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_60": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_5": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_14": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_5": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_61": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_16": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_62": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_34": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_63": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_64": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_65": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_41": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_42": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_43": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_6": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_89": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_90": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_11": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_66": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_30": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_86": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_87": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_12": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_67": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_31": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_88": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_91": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_92": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_93": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_94": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_95": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_44": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_45": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_96": { + "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 + }, + "_to_copy_97": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_90": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_98": { + "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 + }, + "_to_copy_99": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_46": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_51": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_69": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_100": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_22": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_101": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_71": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_93": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_52": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_18": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_53": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_94": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_54": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_72": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_102": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_103": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_74": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_96": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_16": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_55": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "view_47": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "slice_68": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_69": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_10": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_unsafe_view_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_48": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_104": { + "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 + }, + "_to_copy_105": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_36": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_56": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_49": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "_to_copy_106": { + "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 + }, + "_to_copy_107": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_37": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "linear_57": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_50": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_38": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_15": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "unsqueeze_16": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_70": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_71": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 12288 + }, + "slice_72": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_73": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 3072 + }, + "mul_101": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_74": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "slice_75": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "neg_8": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "cat_11": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_102": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "add_79": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_103": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_76": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_77": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "neg_9": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "cat_12": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "mul_104": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "add_80": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "cat_13": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "cat_14": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "cache_update_2": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_3": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_17": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_4": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "unsqueeze_18": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_5": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "transpose_39": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "matmul_3": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "mul_105": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "add_81": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "softmax_1": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_to_copy_108": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "matmul_4": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_40": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_13": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_51": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_7": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_106": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_58": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_82": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_109": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_26": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_110": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_84": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_108": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_19": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_59": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_19": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_60": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_109": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_61": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_85": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_111": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_27": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_112": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_87": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_111": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_20": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_62": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_41": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_63": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_52": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_64": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_65": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_15": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_78": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_6": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_19": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_6": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_79": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_20": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_80": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_42": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_81": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_82": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_83": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_53": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_54": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_55": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_8": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_113": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_114": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_13": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_89": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_39": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_114": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_115": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_14": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_90": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_40": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_116": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_115": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_116": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_117": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_118": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_119": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_56": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_57": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_120": { + "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 + }, + "_to_copy_121": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_118": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_122": { + "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 + }, + "_to_copy_123": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_58": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_66": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_92": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_124": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_29": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_125": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_94": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_121": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_21": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_67": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_22": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_68": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_122": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_69": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_95": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_126": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_30": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_127": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_97": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_124": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_22": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_70": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_71": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_59": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_72": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_73": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_16": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_86": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_7": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_20": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_7": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_87": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_23": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_88": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_45": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_89": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_90": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_91": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_60": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_61": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_62": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_9": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_128": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_129": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_15": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_99": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_44": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_127": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_128": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_16": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_100": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_45": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_129": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_130": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_131": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_132": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_133": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_134": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_63": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_64": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_135": { + "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 + }, + "_to_copy_136": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_131": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_137": { + "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 + }, + "_to_copy_138": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_65": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_74": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_102": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_139": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_32": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_140": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_104": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_134": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_75": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_25": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_76": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_135": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_77": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_105": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_141": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_33": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_142": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_107": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_137": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_24": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_78": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_47": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_79": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_66": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_80": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_81": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_17": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_94": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_8": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_21": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_8": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_95": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_26": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_96": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_48": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_97": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_98": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_99": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_67": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_68": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_69": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_10": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_143": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_8": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_144": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_17": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_109": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_49": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_140": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_141": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_18": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_110": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_50": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_142": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_145": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_146": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_147": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_148": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_149": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_70": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_71": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_150": { + "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 + }, + "_to_copy_151": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_144": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_152": { + "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 + }, + "_to_copy_153": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_72": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_82": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_112": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_154": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_35": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_155": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_114": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_147": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_25": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_83": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_28": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_84": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_148": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_85": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_115": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_156": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_36": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_157": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_117": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_150": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_26": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_86": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "view_73": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "slice_102": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_103": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_17": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_unsafe_view_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_74": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_158": { + "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 + }, + "_to_copy_159": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_50": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_87": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_75": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "_to_copy_160": { + "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 + }, + "_to_copy_161": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_51": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "linear_88": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_76": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_52": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_22": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "unsqueeze_23": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_104": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_105": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 12288 + }, + "slice_106": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_107": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 3072 + }, + "mul_155": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_108": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "slice_109": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "neg_13": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "cat_18": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_156": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "add_122": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_157": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_110": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_111": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "neg_14": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "cat_19": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "mul_158": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "add_123": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "cat_20": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "cat_21": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "cache_update_4": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_5": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_24": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_6": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_18": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_7": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "unsqueeze_25": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_7": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_19": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_8": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "transpose_53": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "matmul_5": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "mul_159": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "add_124": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "softmax_2": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_to_copy_162": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "matmul_6": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_54": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_20": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_77": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_11": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_160": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_89": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_125": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_163": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_39": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_164": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_127": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_162": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_29": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_90": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_29": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_91": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_163": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_92": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_128": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_165": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_40": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_166": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_130": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_165": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_30": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_93": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_55": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_94": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_78": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_95": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_96": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_22": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_112": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_9": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_26": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_9": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_113": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_30": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_114": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_56": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_115": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_116": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_117": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_79": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_80": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_81": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_12": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_167": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_9": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_168": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_19": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_132": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_58": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_168": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_169": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_20": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_133": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_59": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_170": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_169": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_170": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_171": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_172": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_173": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_82": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_83": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_174": { + "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 + }, + "_to_copy_175": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_172": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_176": { + "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 + }, + "_to_copy_177": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_84": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_97": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_135": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_178": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_42": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_179": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_137": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_175": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_31": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_98": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_32": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_99": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_176": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_100": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_138": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_180": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_43": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_181": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_140": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_178": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_32": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_101": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_58": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_102": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_85": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_103": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_104": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_23": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_120": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_10": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_27": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_10": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_121": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_33": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_122": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_59": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_123": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_124": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_125": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_86": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_87": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_88": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_13": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_182": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_183": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_21": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_142": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_63": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_181": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_182": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_22": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_143": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_64": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_183": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_184": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_185": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_186": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_187": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_188": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_89": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_90": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_189": { + "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 + }, + "_to_copy_190": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_185": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_191": { + "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 + }, + "_to_copy_192": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_91": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_105": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_145": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_193": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_45": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_194": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_147": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_188": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_33": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_106": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_35": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_107": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_189": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_108": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_148": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_195": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_46": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_196": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_150": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_191": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_34": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_109": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_61": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_110": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_92": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_111": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_112": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_24": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_128": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_11": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_28": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_11": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_129": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_36": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_130": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_62": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_131": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_132": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_133": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_93": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_94": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_95": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_14": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_197": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_198": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_23": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_152": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_68": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_194": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_195": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_24": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_153": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_69": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_196": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_199": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_200": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_201": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_202": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_203": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_96": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_97": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_204": { + "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 + }, + "_to_copy_205": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_198": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_206": { + "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 + }, + "_to_copy_207": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_98": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_113": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_155": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_208": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_48": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_209": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_157": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_201": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_35": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_114": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_38": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_115": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_202": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_116": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_158": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_210": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_49": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_211": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_160": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_204": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_36": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_117": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "view_99": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "slice_136": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_137": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_24": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_unsafe_view_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_100": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_212": { + "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 + }, + "_to_copy_213": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_64": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_118": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_101": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "_to_copy_214": { + "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 + }, + "_to_copy_215": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_65": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "linear_119": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_102": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_66": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_29": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "unsqueeze_30": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_138": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_139": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 12288 + }, + "slice_140": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_141": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 3072 + }, + "mul_209": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_142": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "slice_143": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "neg_18": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "cat_25": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_210": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "add_165": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_211": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_144": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_145": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "neg_19": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "cat_26": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "mul_212": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "add_166": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "cat_27": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "cat_28": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "cache_update_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_31": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_8": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_25": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_10": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "unsqueeze_32": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_9": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_26": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_11": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "transpose_67": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "matmul_7": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "mul_213": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "add_167": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "softmax_3": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_to_copy_216": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "matmul_8": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_68": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_27": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_103": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_214": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_120": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_168": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_217": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_52": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_218": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_170": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_216": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_39": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_121": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_39": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_122": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_217": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_123": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_171": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_219": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_53": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_220": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_173": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_219": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_40": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_124": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_69": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_125": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_104": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_126": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_127": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_29": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_146": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_12": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_33": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_12": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_147": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_40": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_148": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_70": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_149": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_150": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_151": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_105": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_106": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_107": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_16": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_221": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_20": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_222": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_25": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_175": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_77": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_222": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_223": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_26": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_176": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_78": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_224": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_223": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_224": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_225": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_226": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_227": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_108": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_109": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_228": { + "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 + }, + "_to_copy_229": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_226": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_230": { + "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 + }, + "_to_copy_231": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_110": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_128": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_178": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_232": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_55": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_233": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_180": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_229": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_41": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_129": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_42": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_130": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_230": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_131": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_181": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_234": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_56": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_235": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_183": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_232": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_42": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_132": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_72": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_133": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_111": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_134": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_135": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_30": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_154": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_13": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_34": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_13": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_155": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_43": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_156": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_73": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_157": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_158": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_159": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_112": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_113": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_114": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_17": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_236": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_13": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_21": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_237": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_27": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_185": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_82": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_235": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_236": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_28": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_186": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_83": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_237": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_238": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_239": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_240": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_241": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_242": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_115": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_116": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_243": { + "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 + }, + "_to_copy_244": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_239": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_245": { + "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 + }, + "_to_copy_246": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_117": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_136": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_188": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_247": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_58": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_248": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_190": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_242": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_43": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_137": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_45": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_138": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_243": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_139": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_191": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_249": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_59": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_250": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_193": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_245": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_44": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_140": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_75": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_141": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_118": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_142": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_143": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_31": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_162": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_14": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_35": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_14": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_163": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_46": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_164": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_76": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_165": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_166": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_167": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_119": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_120": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_121": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_18": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_251": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_14": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_22": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_252": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_29": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_195": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_87": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_248": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_249": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_30": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_196": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_88": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_250": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_253": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_254": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_255": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_256": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_257": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_122": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_123": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_258": { + "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 + }, + "_to_copy_259": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_252": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_260": { + "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 + }, + "_to_copy_261": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_124": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_144": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_198": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_262": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_61": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_263": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_200": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_255": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_45": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_145": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_48": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_146": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_256": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_147": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_201": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_264": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_62": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_265": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_203": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_258": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_46": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_148": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "view_125": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "slice_170": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_171": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_31": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_unsafe_view_12": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_126": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_266": { + "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 + }, + "_to_copy_267": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_78": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_149": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_127": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "_to_copy_268": { + "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 + }, + "_to_copy_269": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_79": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "linear_150": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_128": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_80": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_36": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "unsqueeze_37": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_172": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_173": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 12288 + }, + "slice_174": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_175": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 3072 + }, + "mul_263": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_176": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "slice_177": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "neg_23": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "cat_32": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_264": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "add_208": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_265": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_178": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_179": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "neg_24": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "cat_33": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "mul_266": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "add_209": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "cat_34": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "cat_35": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "cache_update_8": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_9": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_38": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_10": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_32": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_13": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "unsqueeze_39": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_33": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_14": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "transpose_81": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "matmul_9": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "mul_267": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "add_210": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "softmax_4": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_to_copy_270": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "matmul_10": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_82": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_34": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_129": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_19": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_268": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_151": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_211": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_271": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_65": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_272": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_213": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_270": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_49": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_152": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_49": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_153": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_271": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_154": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_214": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_273": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_66": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_274": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_216": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_273": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_50": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_155": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_83": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_156": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_130": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_157": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_158": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_36": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_180": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_15": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_40": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_15": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_181": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_50": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_182": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_84": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_183": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_184": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_185": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_131": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_132": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_133": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_20": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_275": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_25": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_276": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_31": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_218": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_96": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_276": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_277": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_32": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_219": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_97": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_278": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_277": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_278": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_279": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_280": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_281": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_134": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_135": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_282": { + "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 + }, + "_to_copy_283": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_280": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_284": { + "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 + }, + "_to_copy_285": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_136": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_159": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_221": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_286": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_68": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_287": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_223": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_283": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_51": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_160": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_52": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_161": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_284": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_162": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_224": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_288": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_69": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_289": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_226": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_286": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_52": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_163": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_86": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_164": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_137": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_165": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_166": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_37": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_188": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_16": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_41": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_16": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_189": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_53": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_190": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_87": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_191": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_192": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_193": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_138": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_139": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_140": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_21": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_290": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_26": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_291": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_33": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_228": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_101": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_289": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_290": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_34": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_229": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_102": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_291": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_292": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_293": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_294": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_295": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_296": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_141": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_142": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_297": { + "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 + }, + "_to_copy_298": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_293": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_299": { + "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 + }, + "_to_copy_300": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_143": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_167": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_231": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_301": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_71": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_302": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_233": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_296": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_53": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_168": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_55": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_169": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_297": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_170": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_234": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_303": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_72": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_304": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_236": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_299": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_54": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_171": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_89": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_172": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_144": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_173": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "linear_174": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "cat_38": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_196": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "squeeze_17": { + "shape": [ + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "unsqueeze_42": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "conv1d_17": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float16", + "role": "value", + "byteLength": 135168 + }, + "slice_197": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "silu_56": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float16", + "role": "value", + "byteLength": 98304 + }, + "slice_198": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "transpose_90": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "slice_199": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_200": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_201": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_145": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_146": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_147": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_22": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "_to_copy_305": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "exp_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_27": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "_to_copy_306": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_35": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_238": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_106": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_302": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_303": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sum_36": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "add_239": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "rsqrt_107": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float16", + "role": "value", + "byteLength": 128 + }, + "mul_304": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_307": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_308": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_309": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_310": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "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 + }, + "_to_copy_311": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_148": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_149": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_312": { + "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 + }, + "_to_copy_313": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_306": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_314": { + "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 + }, + "_to_copy_315": { + "shape": [ + 64, + 128 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_150": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_175": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_241": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_316": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_74": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_317": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_243": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_309": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_55": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_176": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_58": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_177": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_310": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_178": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_244": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_318": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_75": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_319": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_246": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_312": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_56": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_179": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "view_151": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 32768 + }, + "slice_204": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "slice_205": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_38": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_unsafe_view_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_152": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_320": { + "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 + }, + "_to_copy_321": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_92": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_180": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_153": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "_to_copy_322": { + "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 + }, + "_to_copy_323": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "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": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_93": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "linear_181": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "view_154": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "transpose_94": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_43": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "unsqueeze_44": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_206": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_207": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 12288 + }, + "slice_208": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_209": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float16", + "role": "value", + "byteLength": 3072 + }, + "mul_317": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "slice_210": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "slice_211": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "neg_28": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 2048 + }, + "cat_39": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_318": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "add_251": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "mul_319": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "slice_212": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "slice_213": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "neg_29": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float16", + "role": "value", + "byteLength": 512 + }, + "cat_40": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "mul_320": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "add_252": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1024 + }, + "cat_41": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "cat_42": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4096 + }, + "cache_update_10": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_11": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_45": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_39": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_16": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "unsqueeze_46": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 4194304 + }, + "expand_13": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "clone_40": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "_unsafe_view_17": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "transpose_95": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16777216 + }, + "matmul_11": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "mul_321": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "add_253": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "softmax_5": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_to_copy_324": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "matmul_12": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "transpose_96": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "clone_41": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "view_155": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "sigmoid_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "mul_322": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_182": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_254": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_325": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_78": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_326": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_256": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_324": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_59": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_183": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "silu_59": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_184": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "mul_325": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float16", + "role": "value", + "byteLength": 49152 + }, + "linear_185": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "add_257": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "_to_copy_327": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_79": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "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, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_to_copy_328": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_259": { + "shape": [ + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_327": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_60": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "alias": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_186": { + "shape": [ + 1, + 4, + 248320 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1986560 + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 22331136, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 28622592, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 34914048, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 41205504, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 47496960, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 53788416, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 60079872, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00049.bin", + "offset": 66371328, + "byteLength": 737536 + }, + { + "path": "weights/00050.bin", + "offset": 0, + "byteLength": 5553920 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 5553920, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 11845376, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 18136832, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 24428288, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 30719744, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 37011200, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 43302656, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 49594112, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 55885568, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00050.bin", + "offset": 62177024, + "byteLength": 4931840 + }, + { + "path": "weights/00051.bin", + "offset": 0, + "byteLength": 1359616 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00051.bin", + "offset": 1359616, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00051.bin", + "offset": 7651072, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00051.bin", + "offset": 13942528, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00051.bin", + "offset": 20233984, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00051.bin", + "offset": 26525440, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": { + "shape": [ + 6144, + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "parts": [ + { + "path": "weights/00051.bin", + "offset": 32816896, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_0_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 0 + }, + "parts": [ + { + "path": "weights/00000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00001.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_0": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_32768_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 134217728 + }, + "parts": [ + { + "path": "weights/00002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00003.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_1": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_65536_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 268435456 + }, + "parts": [ + { + "path": "weights/00004.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00005.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_2": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_98304_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 402653184 + }, + "parts": [ + { + "path": "weights/00006.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00007.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_131072_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 536870912 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00009.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_4": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_163840_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 671088640 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00011.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_5": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_196608_32768": { + "shape": [ + 32768, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 805306368 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00013.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "embedding__part_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "p_model_model_language_model_embed_tokens_weight__rows_229376_18944": { + "shape": [ + 18944, + 2048 + ], + "dtype": "float16", + "role": "constant", + "byteLength": 77594624, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 939524096 + }, + "parts": [ + { + "path": "weights/00014.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/00015.bin", + "offset": 0, + "byteLength": 10485760 + } + ] + }, + "embedding__part_7": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_0": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_1": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_2": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_4": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_5": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float16", + "role": "value", + "byteLength": 16384 + }, + "linear_186__part_0": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_1": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_2": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_3": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_4": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_5": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_6": { + "shape": [ + 1, + 4, + 32768 + ], + "dtype": "float16", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_7": { + "shape": [ + 1, + 4, + 18944 + ], + "dtype": "float16", + "role": "value", + "byteLength": 151552 + }, + "linear_186__cat_0_0": { + "shape": [ + 1, + 4, + 229376 + ], + "dtype": "float16", + "role": "value", + "byteLength": 1835008 + }, + "linear_186__cat_0_7": { + "shape": [ + 1, + 4, + 18944 + ], + "dtype": "float16", + "role": "value", + "byteLength": 151552 + } + }, + "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": "_to_copy_1", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \".49\", line 37, in forward\n to_5 = torch.ops.aten.to.dtype(mul, torch.float16); mul = None" + }, + { + "id": "_to_copy_2", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_1" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \".49\", line 39, in forward\n to_6 = torch.ops.aten.to.dtype(mul_1, torch.float16); mul_1 = None" + }, + { + "id": "_to_copy_3", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "embedding" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_1", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_3" + }, + { + "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": "_to_copy_4", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_1", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_4" + }, + 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": "_to_copy_5", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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": "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": "_to_copy_6", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_3" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_6" + }, + { + "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": "_to_copy_7", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_6" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_8", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_8" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_9", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_3" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_10", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_7" + }, + { + "tensor": "_to_copy_8" + }, + { + "tensor": "_to_copy_9" + }, + { + "tensor": "mul_4" + }, + { + "tensor": "_to_copy_10" + }, + { + "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": "_to_copy_11", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_11" + }, + [ + -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": "_to_copy_12", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_4" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_2", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_12" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_12" + }, + { + "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": "_to_copy_13", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_9" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_10", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_14", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_5" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_1", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\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": "_to_copy_15", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_11" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_6", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_15" + }, + [ + 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": "_to_copy_16", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_6" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_3", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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_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": "_to_copy_16" + }, + { + "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": "_to_copy_17", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_8", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_17" + }, + 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": "_to_copy_18", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_9" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_4", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_18" + }, + { + "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": "_to_copy_19", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_11", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_19" + }, + 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": "_to_copy_20", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_1", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_21", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_11" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_21" + }, + { + "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": "_to_copy_22", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_19" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_23", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_21" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_24", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_10" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_25", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_1" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_1", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_22" + }, + { + "tensor": "_to_copy_23" + }, + { + "tensor": "_to_copy_24" + }, + { + "tensor": "mul_17" + }, + { + "tensor": "_to_copy_25" + }, + { + "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": "_to_copy_26", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_1" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_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 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": "_to_copy_27", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_11" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_5", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_27" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_27" + }, + { + "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": "_to_copy_28", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_22" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_23", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_29", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_12" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_4", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_30", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_24" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_13", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_30" + }, + [ + 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": "_to_copy_31", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_16" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_6", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_31" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_31" + }, + { + "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": "_to_copy_32", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_18", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_32" + }, + 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": "_to_copy_33", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_19" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_7", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_33" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_33" + }, + { + "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": "_to_copy_34", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_21", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_34" + }, + 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": "_to_copy_35", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_2", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_36", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_19" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_36" + }, + { + "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": "_to_copy_37", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_32" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_38", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_34" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_39", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_17" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_40", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_2" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_2", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_37" + }, + { + "tensor": "_to_copy_38" + }, + { + "tensor": "_to_copy_39" + }, + { + "tensor": "mul_30" + }, + { + "tensor": "_to_copy_40" + }, + { + "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": "_to_copy_41", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_2" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_41" + }, + [ + -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": "_to_copy_42", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_18" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_8", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_42" + }, + { + "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": "_to_copy_43", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_35" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_36", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_44", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_19" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_7", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\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": "_to_copy_45", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_37" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_20", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_45" + }, + [ + 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": "_to_copy_46", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_26" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_9", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_46" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_46" + }, + { + "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": "_to_copy_47", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_28", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_47" + }, + 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": "_to_copy_48", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_29" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_10", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_48" + }, + { + "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": "_to_copy_49", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_31", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_49" + }, + 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": "_to_copy_50", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_22" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_11", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_50" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_50" + }, + { + "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": "_to_copy_51", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_33", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_51" + }, + 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": "_to_copy_52", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_23" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_12", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.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": "_to_copy_52" + }, + { + "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": "_to_copy_53", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_35", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_53" + }, + 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": "_to_copy_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 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_9", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, 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": "_to_copy_54", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "softmax" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = 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": "_to_copy_54" + }, + { + "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": "_to_copy_55", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_39" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_13", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_55" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_55" + }, + { + "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": "_to_copy_56", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_41", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_56" + }, + 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": "_to_copy_57", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_42" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_14", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_57" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_57" + }, + { + "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": "_to_copy_58", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_44", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_58" + }, + 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": "_to_copy_59", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_3", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_60", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_34" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_60" + }, + { + "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": "_to_copy_61", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_60" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_62", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_62" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_63", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_29" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_64", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_4" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_3", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_61" + }, + { + "tensor": "_to_copy_62" + }, + { + "tensor": "_to_copy_63" + }, + { + "tensor": "mul_58" + }, + { + "tensor": "_to_copy_64" + }, + { + "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": "_to_copy_65", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_7" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_65" + }, + [ + -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": "_to_copy_66", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_30" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_15", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_66" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_66" + }, + { + "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": "_to_copy_67", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_63" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_64", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_68", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_31" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_11", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_69", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_65" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_32", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_69" + }, + [ + 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": "_to_copy_70", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_49" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_16", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_70" + }, + { + "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": "_to_copy_71", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_51", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_71" + }, + 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": "_to_copy_72", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_52" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_17", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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_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": "_to_copy_72" + }, + { + "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": "_to_copy_73", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_54", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_73" + }, + 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": "_to_copy_74", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_4", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_75", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_42" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_75" + }, + { + "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": "_to_copy_76", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_73" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_77", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_75" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_78", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_36" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_79", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_5" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_4", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_76" + }, + { + "tensor": "_to_copy_77" + }, + { + "tensor": "_to_copy_78" + }, + { + "tensor": "mul_71" + }, + { + "tensor": "_to_copy_79" + }, + { + "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": "_to_copy_80", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_8" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_80" + }, + [ + -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": "_to_copy_81", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_37" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_18", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_81" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_81" + }, + { + "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": "_to_copy_82", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_76" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_77", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_83", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_38" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_14", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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_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": "_to_copy_84", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_78" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_39", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_84" + }, + [ + 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": "_to_copy_85", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_59" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_19", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_85" + }, + { + "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": "_to_copy_86", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_61", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_86" + }, + 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": "_to_copy_87", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_62" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_20", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_87" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_87" + }, + { + "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": "_to_copy_88", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_64", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_88" + }, + 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": "_to_copy_89", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_5", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_90", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_50" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_90" + }, + { + "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": "_to_copy_91", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_86" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_92", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_88" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_93", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_43" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_94", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_6" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_5", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_91" + }, + { + "tensor": "_to_copy_92" + }, + { + "tensor": "_to_copy_93" + }, + { + "tensor": "mul_84" + }, + { + "tensor": "_to_copy_94" + }, + { + "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": "_to_copy_95", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_9" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_95" + }, + [ + -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": "_to_copy_96", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_44" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_21", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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_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": "_to_copy_96" + }, + { + "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": "_to_copy_97", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_89" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_90", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_98", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_45" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_17", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_99", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_91" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_46", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_99" + }, + [ + 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": "_to_copy_100", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_69" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_22", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_100" + }, + { + "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": "_to_copy_101", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_71", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_101" + }, + 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": "_to_copy_102", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_72" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_23", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_102" + }, + { + "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": "_to_copy_103", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_74", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_103" + }, + 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": "_to_copy_104", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_48" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_24", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_104" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_104" + }, + { + "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": "_to_copy_105", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_76", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_105" + }, + 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": "_to_copy_106", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_49" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_25", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_106" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_106" + }, + { + "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": "_to_copy_107", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_78", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_107" + }, + 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": "_to_copy_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 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_16", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, 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": "_to_copy_108", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "softmax_1" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = 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": "_to_copy_108" + }, + { + "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": "_to_copy_109", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_82" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_26", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_109" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_109" + }, + { + "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": "_to_copy_110", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_84", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_110" + }, + 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": "_to_copy_111", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_85" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_27", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_111" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_111" + }, + { + "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": "_to_copy_112", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_87", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_112" + }, + 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": "_to_copy_113", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_6", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_114", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_65" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_114" + }, + { + "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": "_to_copy_115", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_114" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_116", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_116" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_117", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_55" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_118", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_8" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_6", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_115" + }, + { + "tensor": "_to_copy_116" + }, + { + "tensor": "_to_copy_117" + }, + { + "tensor": "mul_112" + }, + { + "tensor": "_to_copy_118" + }, + { + "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": "_to_copy_119", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_14" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_119" + }, + [ + -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": "_to_copy_120", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_56" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_28", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_120" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_120" + }, + { + "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": "_to_copy_121", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_117" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_118", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_122", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_57" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_21", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_123", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_119" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_58", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_123" + }, + [ + 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": "_to_copy_124", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_92" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_29", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_124" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_124" + }, + { + "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": "_to_copy_125", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_94", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_125" + }, + 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": "_to_copy_126", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_95" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_30", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_126" + }, + { + "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": "_to_copy_127", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_97", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_127" + }, + 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": "_to_copy_128", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_7", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_129", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_73" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_129" + }, + { + "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": "_to_copy_130", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_127" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_131", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_129" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_132", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_62" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_133", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_9" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_7", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_130" + }, + { + "tensor": "_to_copy_131" + }, + { + "tensor": "_to_copy_132" + }, + { + "tensor": "mul_125" + }, + { + "tensor": "_to_copy_133" + }, + { + "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": "_to_copy_134", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_15" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_134" + }, + [ + -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": "_to_copy_135", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_63" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_31", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_135" + }, + { + "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": "_to_copy_136", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_130" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_131", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_137", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_64" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_24", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_138", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_132" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_65", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_138" + }, + [ + 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": "_to_copy_139", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_102" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_32", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_139" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_139" + }, + { + "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": "_to_copy_140", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_104", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_140" + }, + 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": "_to_copy_141", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_105" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_33", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_141" + }, + { + "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": "_to_copy_142", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_107", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_142" + }, + 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": "_to_copy_143", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_8", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_144", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_81" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_144" + }, + { + "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": "_to_copy_145", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_140" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_146", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_142" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_147", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_69" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_148", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_10" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_8", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_145" + }, + { + "tensor": "_to_copy_146" + }, + { + "tensor": "_to_copy_147" + }, + { + "tensor": "mul_138" + }, + { + "tensor": "_to_copy_148" + }, + { + "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": "_to_copy_149", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_16" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_149" + }, + [ + -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": "_to_copy_150", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_70" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_34", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_150" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_150" + }, + { + "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": "_to_copy_151", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_143" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_144", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_152", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_71" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_27", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_153", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_145" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_72", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_153" + }, + [ + 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": "_to_copy_154", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_112" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_35", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_154" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_154" + }, + { + "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": "_to_copy_155", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_114", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_155" + }, + 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": "_to_copy_156", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_115" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_36", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_156" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_156" + }, + { + "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": "_to_copy_157", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_117", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_157" + }, + 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": "_to_copy_158", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_74" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_37", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.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": "_to_copy_158" + }, + { + "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": "_to_copy_159", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_119", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_159" + }, + 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": "_to_copy_160", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_75" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_38", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_160" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_160" + }, + { + "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": "_to_copy_161", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_121", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_161" + }, + 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": "_to_copy_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 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_23", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, 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": "_to_copy_162", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "softmax_2" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = 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": "_to_copy_162" + }, + { + "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": "_to_copy_163", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_125" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_39", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_163" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_163" + }, + { + "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": "_to_copy_164", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_127", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_164" + }, + 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": "_to_copy_165", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_128" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_40", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_165" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_165" + }, + { + "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": "_to_copy_166", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_130", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_166" + }, + 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": "_to_copy_167", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_9", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_168", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_96" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_168" + }, + { + "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": "_to_copy_169", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_168" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_170", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_170" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_171", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_81" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_172", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_12" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_9", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_169" + }, + { + "tensor": "_to_copy_170" + }, + { + "tensor": "_to_copy_171" + }, + { + "tensor": "mul_166" + }, + { + "tensor": "_to_copy_172" + }, + { + "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": "_to_copy_173", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_21" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_173" + }, + [ + -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": "_to_copy_174", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_82" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_41", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_174" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_174" + }, + { + "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": "_to_copy_175", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_171" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_172", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_176", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_83" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_31", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_177", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_173" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_84", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_177" + }, + [ + 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": "_to_copy_178", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_135" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_42", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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_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": "_to_copy_178" + }, + { + "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": "_to_copy_179", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_137", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_179" + }, + 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": "_to_copy_180", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_138" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_43", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_180" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_180" + }, + { + "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": "_to_copy_181", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_140", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_181" + }, + 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": "_to_copy_182", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_10", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_183", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_104" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_183" + }, + { + "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": "_to_copy_184", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_181" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_185", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_183" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_186", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_88" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_187", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_13" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_10", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_184" + }, + { + "tensor": "_to_copy_185" + }, + { + "tensor": "_to_copy_186" + }, + { + "tensor": "mul_179" + }, + { + "tensor": "_to_copy_187" + }, + { + "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": "_to_copy_188", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_22" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_188" + }, + [ + -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": "_to_copy_189", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_89" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_44", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_189" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_189" + }, + { + "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": "_to_copy_190", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_184" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_185", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_191", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_90" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_34", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_192", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_186" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_91", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_192" + }, + [ + 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": "_to_copy_193", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_145" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_45", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_193" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_193" + }, + { + "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": "_to_copy_194", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_147", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_194" + }, + 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": "_to_copy_195", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_148" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_46", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_195" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_195" + }, + { + "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": "_to_copy_196", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_150", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_196" + }, + 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": "_to_copy_197", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_11", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_198", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_112" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_198" + }, + { + "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": "_to_copy_199", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_194" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_200", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_196" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_201", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_95" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_202", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_14" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_11", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_199" + }, + { + "tensor": "_to_copy_200" + }, + { + "tensor": "_to_copy_201" + }, + { + "tensor": "mul_192" + }, + { + "tensor": "_to_copy_202" + }, + { + "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": "_to_copy_203", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_23" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_203" + }, + [ + -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": "_to_copy_204", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_96" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_47", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_204" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_204" + }, + { + "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": "_to_copy_205", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_197" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_198", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_206", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_97" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_37", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_207", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_199" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_98", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_207" + }, + [ + 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": "_to_copy_208", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_155" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_48", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_208" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_208" + }, + { + "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": "_to_copy_209", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_157", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_209" + }, + 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": "_to_copy_210", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_158" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_49", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_210" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_210" + }, + { + "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": "_to_copy_211", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_160", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_211" + }, + 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": "_to_copy_212", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_100" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_50", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_212" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_212" + }, + { + "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": "_to_copy_213", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_162", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_213" + }, + 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": "_to_copy_214", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_101" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_51", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.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": "_to_copy_214" + }, + { + "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": "_to_copy_215", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_164", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_215" + }, + 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": "_to_copy_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 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_30", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, 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": "_to_copy_216", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "softmax_3" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = 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": "_to_copy_216" + }, + { + "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": "_to_copy_217", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_168" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_52", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_217" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_217" + }, + { + "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": "_to_copy_218", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_170", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_218" + }, + 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": "_to_copy_219", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_171" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_53", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_219" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_219" + }, + { + "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": "_to_copy_220", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_173", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_220" + }, + 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": "_to_copy_221", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_12", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_222", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_127" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_222" + }, + { + "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": "_to_copy_223", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_222" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_224", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_224" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_225", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_107" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_226", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_16" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_12", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_223" + }, + { + "tensor": "_to_copy_224" + }, + { + "tensor": "_to_copy_225" + }, + { + "tensor": "mul_220" + }, + { + "tensor": "_to_copy_226" + }, + { + "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": "_to_copy_227", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_28" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_227" + }, + [ + -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": "_to_copy_228", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_108" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_54", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_228" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_228" + }, + { + "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": "_to_copy_229", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_225" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_226", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_230", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_109" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_41", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_231", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_227" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_110", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_231" + }, + [ + 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": "_to_copy_232", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_178" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_55", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_232" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_232" + }, + { + "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": "_to_copy_233", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_180", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_233" + }, + 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": "_to_copy_234", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_181" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_56", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_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_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": "_to_copy_234" + }, + { + "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": "_to_copy_235", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_183", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_235" + }, + 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": "_to_copy_236", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_13", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_237", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_135" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_237" + }, + { + "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": "_to_copy_238", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_235" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_239", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_237" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_240", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_114" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_241", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_17" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_13", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_238" + }, + { + "tensor": "_to_copy_239" + }, + { + "tensor": "_to_copy_240" + }, + { + "tensor": "mul_233" + }, + { + "tensor": "_to_copy_241" + }, + { + "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": "_to_copy_242", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_29" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_242" + }, + [ + -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": "_to_copy_243", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_115" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_57", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_243" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_243" + }, + { + "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": "_to_copy_244", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_238" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_239", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_245", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_116" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_44", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_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 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_246", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_240" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_117", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_246" + }, + [ + 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": "_to_copy_247", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_188" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_58", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_247" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_247" + }, + { + "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": "_to_copy_248", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_190", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_248" + }, + 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": "_to_copy_249", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_191" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_59", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_249" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_249" + }, + { + "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": "_to_copy_250", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_193", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_250" + }, + 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": "_to_copy_251", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_14", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_252", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_143" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_252" + }, + { + "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": "_to_copy_253", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_248" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_254", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_250" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_255", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_121" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_256", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_18" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_14", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_253" + }, + { + "tensor": "_to_copy_254" + }, + { + "tensor": "_to_copy_255" + }, + { + "tensor": "mul_246" + }, + { + "tensor": "_to_copy_256" + }, + { + "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": "_to_copy_257", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_30" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_257" + }, + [ + -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": "_to_copy_258", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_122" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_60", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_258" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_258" + }, + { + "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": "_to_copy_259", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_251" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_252", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_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 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_260", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_123" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_47", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_260" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_261", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_253" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_124", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_261" + }, + [ + 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": "_to_copy_262", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_198" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_61", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_262" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_262" + }, + { + "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": "_to_copy_263", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_200", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_263" + }, + 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": "_to_copy_264", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_201" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_62", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_264" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_264" + }, + { + "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": "_to_copy_265", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_203", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_265" + }, + 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": "_to_copy_266", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_126" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_63", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_266" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_266" + }, + { + "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": "_to_copy_267", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_205", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_267" + }, + 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": "_to_copy_268", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_127" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_64", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_268" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_268" + }, + { + "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": "_to_copy_269", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_207", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_269" + }, + 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": "_to_copy_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 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_37", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, 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": "_to_copy_270", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "softmax_4" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = 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": "_to_copy_270" + }, + { + "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": "_to_copy_271", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_211" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_65", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_271" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_271" + }, + { + "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": "_to_copy_272", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_213", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_272" + }, + 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": "_to_copy_273", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_214" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_66", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_273" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_273" + }, + { + "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": "_to_copy_274", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_216", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_274" + }, + 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": "_to_copy_275", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_15", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_275" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_276", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_158" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_276" + }, + { + "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": "_to_copy_277", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_276" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_278", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_278" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_279", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_133" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_280", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_20" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_15", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_277" + }, + { + "tensor": "_to_copy_278" + }, + { + "tensor": "_to_copy_279" + }, + { + "tensor": "mul_274" + }, + { + "tensor": "_to_copy_280" + }, + { + "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": "_to_copy_281", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_35" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_281" + }, + [ + -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": "_to_copy_282", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_134" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_67", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_282" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_282" + }, + { + "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": "_to_copy_283", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_279" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_280", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_283" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_284", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_135" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_51", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_284" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_285", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_281" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_136", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_285" + }, + [ + 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": "_to_copy_286", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_221" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_68", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_286" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_286" + }, + { + "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": "_to_copy_287", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_223", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_287" + }, + 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": "_to_copy_288", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_224" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_69", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_288" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_288" + }, + { + "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": "_to_copy_289", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_226", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_289" + }, + 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": "_to_copy_290", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_16", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_290" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_291", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_166" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_291" + }, + { + "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": "_to_copy_292", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_289" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_293", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_291" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_294", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_140" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_295", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_21" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_16", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_292" + }, + { + "tensor": "_to_copy_293" + }, + { + "tensor": "_to_copy_294" + }, + { + "tensor": "mul_287" + }, + { + "tensor": "_to_copy_295" + }, + { + "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": "_to_copy_296", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_36" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_296" + }, + [ + -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": "_to_copy_297", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_141" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_70", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_297" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_297" + }, + { + "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": "_to_copy_298", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_292" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_293", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_298" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_299", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_142" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_54", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_299" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_300", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_294" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_143", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_300" + }, + [ + 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": "_to_copy_301", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_231" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_71", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_301" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_301" + }, + { + "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": "_to_copy_302", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_233", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_302" + }, + 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": "_to_copy_303", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_234" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_72", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_303" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_303" + }, + { + "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": "_to_copy_304", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_236", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_304" + }, + 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": "_to_copy_305", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_a_log" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "exp_17", + "op": "aten.exp.default", + "args": [ + { + "tensor": "_to_copy_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 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": "_to_copy_306", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "linear_174" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_306" + }, + { + "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": "_to_copy_307", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_302" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()" + }, + { + "id": "_to_copy_308", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_304" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_309", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_147" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_to_copy_310", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "sigmoid_22" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 29, in _webtorch_chunk_gated_delta_rule\n query.float(), key.float(), value.float(), g.float(), beta.float(), initial_state.float()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_17", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "_to_copy_307" + }, + { + "tensor": "_to_copy_308" + }, + { + "tensor": "_to_copy_309" + }, + { + "tensor": "mul_300" + }, + { + "tensor": "_to_copy_310" + }, + { + "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": "_to_copy_311", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "clone_37" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n 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": "_to_copy_311" + }, + [ + -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": "_to_copy_312", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_148" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 177, in forward\n hidden_states = hidden_states.to(torch.float32)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_73", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_312" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_312" + }, + { + "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": "_to_copy_313", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_305" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "mul_306", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_norm_weight" + }, + { + "tensor": "_to_copy_313" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_314", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_149" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "silu_57", + "op": "aten.silu.default", + "args": [ + { + "tensor": "_to_copy_314" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_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": "_to_copy_315", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "mul_307" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 184, in forward\n return hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_150", + "op": "aten.view.default", + "args": [ + { + "tensor": "_to_copy_315" + }, + [ + 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": "_to_copy_316", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_241" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_74", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_316" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_316" + }, + { + "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": "_to_copy_317", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_243", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_317" + }, + 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": "_to_copy_318", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_244" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_75", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_318" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_318" + }, + { + "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": "_to_copy_319", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_input_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_246", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_319" + }, + 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": "_to_copy_320", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_152" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_76", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_320" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_320" + }, + { + "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": "_to_copy_321", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_248", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_321" + }, + 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": "_to_copy_322", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "view_153" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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 ^^^^^^^^^^" + }, + { + "id": "pow_77", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_322" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "_to_copy_322" + }, + { + "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": "_to_copy_323", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.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": "add_250", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_323" + }, + 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": "_to_copy_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 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_44", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "_to_copy_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 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, 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": "_to_copy_324", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "softmax_5" + } + ], + "kwargs": { + "dtype": { + "dtype": "float16" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = 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": "_to_copy_324" + }, + { + "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": "_to_copy_325", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_254" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^" + }, + { + "id": "pow_78", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_325" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/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": "_to_copy_325" + }, + { + "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": "_to_copy_326", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_256", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_326" + }, + 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": "_to_copy_327", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "add_257" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *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 ^^^^^^^^^^" + }, + { + "id": "pow_79", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "_to_copy_327" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *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": "_to_copy_327" + }, + { + "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": "_to_copy_328", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "p_model_model_language_model_norm_weight" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *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": "add_259", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "_to_copy_328" + }, + 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_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.012451171875, + "meanSquaredWeightError": 3.6705523598357104e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0203857421875, + "meanSquaredWeightError": 4.564719517929916e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.02197265625, + "meanSquaredWeightError": 4.645992248697439e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.021331787109375, + "meanSquaredWeightError": 4.5774649493068864e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0203857421875, + "meanSquaredWeightError": 4.7111362277973967e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.024169921875, + "meanSquaredWeightError": 4.975582328370365e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.021636962890625, + "meanSquaredWeightError": 5.357953227758117e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.019012451171875, + "meanSquaredWeightError": 5.218458909439505e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.018585205078125, + "meanSquaredWeightError": 5.083170435682405e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0203094482421875, + "meanSquaredWeightError": 5.197592258809891e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.02294921875, + "meanSquaredWeightError": 5.36857783117739e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.016845703125, + "meanSquaredWeightError": 3.5742041859521123e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.02197265625, + "meanSquaredWeightError": 5.611909159597417e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.021240234375, + "meanSquaredWeightError": 5.885207201572484e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0279541015625, + "meanSquaredWeightError": 6.211186587279371e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.032989501953125, + "meanSquaredWeightError": 7.473293521798041e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0166015625, + "meanSquaredWeightError": 3.48117595194708e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.01544189453125, + "meanSquaredWeightError": 3.525601925957744e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0191650390625, + "meanSquaredWeightError": 3.658423111119191e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.017852783203125, + "meanSquaredWeightError": 3.842288833766361e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0272216796875, + "meanSquaredWeightError": 4.1231859881918353e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.020263671875, + "meanSquaredWeightError": 4.328698253175389e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0208740234375, + "meanSquaredWeightError": 4.316705428664136e-07, + "storageBytes": 12582912 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 12582912, + "groupSize": 8, + "maxAbsoluteWeightError": 0.0255126953125, + "meanSquaredWeightError": 4.453344502053369e-07, + "storageBytes": 12582912 + } + ], + "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_32768" + ], + "output": "embedding__part_0", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/03d3272d16bc278fcc977e95fcb512bac183f503c7d6ced229f9f01f8e5011bf.wgsl" + }, + { + "node": "embedding__part_1", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_32768_32768" + ], + "output": "embedding__part_1", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/16757bdfacc94a92cfc58c9a3ba83ca377b8ea7ddf82fa8f6611f797a6934f14.wgsl" + }, + { + "node": "embedding__part_2", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_65536_32768" + ], + "output": "embedding__part_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b17aaca06f3b631ebb0a4e873183966913175acb34bb0cec516c8bbec20c22dd.wgsl" + }, + { + "node": "embedding__part_3", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_98304_32768" + ], + "output": "embedding__part_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/506e74c3370bbc59adb7e554bbb95deaee0213f0dd19a4942ecb9a36f5dde1a4.wgsl" + }, + { + "node": "embedding__part_4", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_131072_32768" + ], + "output": "embedding__part_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0ecd1d59b46e658df21ec09d4256b75bf9a57138f7bfe5e2bcc309a352a37468.wgsl" + }, + { + "node": "embedding__part_5", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_163840_32768" + ], + "output": "embedding__part_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c5381a1922110bc02e83d499e6af952b8ee49786c658a3666896584b9d89a365.wgsl" + }, + { + "node": "embedding__part_6", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_196608_32768" + ], + "output": "embedding__part_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/58ef8a5a09de45f92e4696b0ce6e26de2500f935d75049d78111d723fdce3780.wgsl" + }, + { + "node": "embedding__part_7", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_229376_18944" + ], + "output": "embedding__part_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f4c3dc2ebe43777fd7845b1e2ffb53880a2262bf3c6626323738f748fe0e679c.wgsl" + }, + { + "node": "embedding__sum_0", + "inputs": [ + "embedding__part_0", + "embedding__part_1" + ], + "output": "embedding__sum_0", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "embedding__sum_1", + "inputs": [ + "embedding__sum_0", + "embedding__part_2" + ], + "output": "embedding__sum_1", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "embedding__sum_2", + "inputs": [ + "embedding__sum_1", + "embedding__part_3" + ], + "output": "embedding__sum_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "embedding__sum_3", + "inputs": [ + "embedding__sum_2", + "embedding__part_4" + ], + "output": "embedding__sum_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "embedding__sum_4", + "inputs": [ + "embedding__sum_3", + "embedding__part_5" + ], + "output": "embedding__sum_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "embedding__sum_5", + "inputs": [ + "embedding__sum_4", + "embedding__part_6" + ], + "output": "embedding__sum_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "embedding", + "inputs": [ + "embedding__sum_5", + "embedding__part_7" + ], + "output": "embedding", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.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": "unsqueeze_1", + "inputs": [ + "b_model_model_language_model_rotary_emb_inv_freq" + ], + "output": "unsqueeze_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c72041e82befee6919af48775291256e22a43dc6445ef0aa288db6b96585a2a9.wgsl" + }, + { + "node": "unsqueeze_2", + "inputs": [ + "unsqueeze_1" + ], + "output": "unsqueeze_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c72041e82befee6919af48775291256e22a43dc6445ef0aa288db6b96585a2a9.wgsl" + }, + { + "node": "unsqueeze_3", + "inputs": [ + "unsqueeze_2" + ], + "output": "unsqueeze_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c72041e82befee6919af48775291256e22a43dc6445ef0aa288db6b96585a2a9.wgsl" + }, + { + "node": "expand_1", + "inputs": [ + "unsqueeze_3" + ], + "output": "expand_1", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.wgsl" + }, + { + "node": "unsqueeze_4", + "inputs": [ + "slice_1" + ], + "output": "unsqueeze_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/82bc226410ee45662442e25040b6a6456c2532b87e96860e10f8f0135dde892b.wgsl" + }, + { + "node": "_to_copy", + "inputs": [ + "unsqueeze_4" + ], + "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": "_to_copy_1", + "inputs": [ + "mul" + ], + "output": "_to_copy_1", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/26655d5856c1c506477f9af344dc55197aa9220fe6e447eb38b11add83e278f3.wgsl" + }, + { + "node": "_to_copy_2", + "inputs": [ + "mul_1" + ], + "output": "_to_copy_2", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/26655d5856c1c506477f9af344dc55197aa9220fe6e447eb38b11add83e278f3.wgsl" + }, + { + "node": "_to_copy_3", + "inputs": [ + "embedding" + ], + "output": "_to_copy_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_1", + "inputs": [ + "_to_copy_3" + ], + "output": "pow_1", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean", + "inputs": [ + "pow_1" + ], + "output": "mean", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add", + "inputs": [ + "mean" + ], + "output": "add", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt", + "inputs": [ + "add" + ], + "output": "rsqrt", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_2", + "inputs": [ + "_to_copy_3", + "rsqrt" + ], + "output": "mul_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_4", + "inputs": [ + "p_model_model_language_model_layers_0_input_layernorm_weight" + ], + "output": "_to_copy_4", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_1", + "inputs": [ + "_to_copy_4" + ], + "output": "add_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_3", + "inputs": [ + "mul_2", + "add_1" + ], + "output": "mul_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as", + "inputs": [ + "mul_3" + ], + "output": "type_as", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear", + "inputs": [ + "type_as", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight" + ], + "output": "linear", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_13", + "inputs": [ + "linear" + ], + "output": "transpose_13", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_1", + "inputs": [ + "type_as", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight" + ], + "output": "linear_1", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view", + "inputs": [ + "linear_1" + ], + "output": "view", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_2", + "inputs": [ + "type_as", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight" + ], + "output": "linear_2", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_3", + "inputs": [ + "type_as", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight" + ], + "output": "linear_3", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_1", + "inputs": [ + "states_0", + "transpose_13" + ], + "output": "cat_1", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_10", + "inputs": [ + "cat_1" + ], + "output": "slice_10", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_0" + }, + { + "node": "squeeze", + "inputs": [ + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + ], + "output": "squeeze", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_5", + "inputs": [ + "squeeze" + ], + "output": "unsqueeze_5", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d", + "inputs": [ + "cat_1", + "unsqueeze_5" + ], + "output": "conv1d", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_11", + "inputs": [ + "conv1d" + ], + "output": "slice_11", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu", + "inputs": [ + "slice_11" + ], + "output": "silu", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_12", + "inputs": [ + "silu" + ], + "output": "slice_12", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_14", + "inputs": [ + "slice_12" + ], + "output": "transpose_14", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_13", + "inputs": [ + "transpose_14" + ], + "output": "slice_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_14", + "inputs": [ + "transpose_14" + ], + "output": "slice_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_15", + "inputs": [ + "transpose_14" + ], + "output": "slice_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_1", + "inputs": [ + "slice_13" + ], + "output": "view_1", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_2", + "inputs": [ + "slice_14" + ], + "output": "view_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_3", + "inputs": [ + "slice_15" + ], + "output": "view_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid", + "inputs": [ + "linear_2" + ], + "output": "sigmoid", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_5", + "inputs": [ + "p_model_model_language_model_layers_0_linear_attn_a_log" + ], + "output": "_to_copy_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp", + "inputs": [ + "_to_copy_5" + ], + "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": "_to_copy_6", + "inputs": [ + "linear_3" + ], + "output": "_to_copy_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_2", + "inputs": [ + "_to_copy_6", + "p_model_model_language_model_layers_0_linear_attn_dt_bias" + ], + "output": "add_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_1" + ], + "output": "mul_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_1", + "inputs": [ + "mul_5" + ], + "output": "sum_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_3", + "inputs": [ + "sum_1" + ], + "output": "add_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_1", + "inputs": [ + "add_3" + ], + "output": "rsqrt_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_6", + "inputs": [ + "view_1", + "rsqrt_1" + ], + "output": "mul_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_7", + "inputs": [ + "view_2" + ], + "output": "mul_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_2", + "inputs": [ + "mul_7" + ], + "output": "sum_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_4", + "inputs": [ + "sum_2" + ], + "output": "add_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_2", + "inputs": [ + "add_4" + ], + "output": "rsqrt_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_8", + "inputs": [ + "view_2", + "rsqrt_2" + ], + "output": "mul_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_7", + "inputs": [ + "mul_6" + ], + "output": "_to_copy_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_8", + "inputs": [ + "mul_8" + ], + "output": "_to_copy_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_9", + "inputs": [ + "view_3" + ], + "output": "_to_copy_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_10", + "inputs": [ + "sigmoid" + ], + "output": "_to_copy_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule", + "inputs": [ + "_to_copy_7", + "_to_copy_8", + "_to_copy_9", + "mul_4", + "_to_copy_10", + "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": "clone", + "inputs": [ + "transpose_15" + ], + "output": "clone", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_11", + "inputs": [ + "clone" + ], + "output": "_to_copy_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_4", + "inputs": [ + "_to_copy_11" + ], + "output": "view_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_5", + "inputs": [ + "view" + ], + "output": "view_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_12", + "inputs": [ + "view_4" + ], + "output": "_to_copy_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_2", + "inputs": [ + "_to_copy_12" + ], + "output": "pow_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_1", + "inputs": [ + "pow_2" + ], + "output": "mean_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_5", + "inputs": [ + "mean_1" + ], + "output": "add_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_3", + "inputs": [ + "add_5" + ], + "output": "rsqrt_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_9", + "inputs": [ + "_to_copy_12", + "rsqrt_3" + ], + "output": "mul_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_13", + "inputs": [ + "mul_9" + ], + "output": "_to_copy_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_10", + "inputs": [ + "p_model_model_language_model_layers_0_linear_attn_norm_weight", + "_to_copy_13" + ], + "output": "mul_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_14", + "inputs": [ + "view_5" + ], + "output": "_to_copy_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_11", + "inputs": [ + "_to_copy_14", + "mul_10" + ], + "output": "mul_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_15", + "inputs": [ + "mul_11" + ], + "output": "_to_copy_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_6", + "inputs": [ + "_to_copy_15" + ], + "output": "view_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_4", + "inputs": [ + "view_6", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight" + ], + "output": "linear_4", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_6", + "inputs": [ + "embedding", + "linear_4" + ], + "output": "add_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_16", + "inputs": [ + "add_6" + ], + "output": "_to_copy_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_3", + "inputs": [ + "_to_copy_16" + ], + "output": "pow_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_2", + "inputs": [ + "pow_3" + ], + "output": "mean_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_7", + "inputs": [ + "mean_2" + ], + "output": "add_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_4", + "inputs": [ + "add_7" + ], + "output": "rsqrt_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_12", + "inputs": [ + "_to_copy_16", + "rsqrt_4" + ], + "output": "mul_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_17", + "inputs": [ + "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + ], + "output": "_to_copy_17", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_8", + "inputs": [ + "_to_copy_17" + ], + "output": "add_8", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_13", + "inputs": [ + "mul_12", + "add_8" + ], + "output": "mul_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_1", + "inputs": [ + "mul_13" + ], + "output": "type_as_1", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_5", + "inputs": [ + "type_as_1", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight" + ], + "output": "linear_5", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_6", + "inputs": [ + "type_as_1", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_14", + "inputs": [ + "linear_5", + "linear_6" + ], + "output": "mul_14", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_7", + "inputs": [ + "mul_14", + "p_model_model_language_model_layers_0_mlp_down_proj_weight" + ], + "output": "linear_7", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_9", + "inputs": [ + "add_6", + "linear_7" + ], + "output": "add_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_18", + "inputs": [ + "add_9" + ], + "output": "_to_copy_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_4", + "inputs": [ + "_to_copy_18" + ], + "output": "pow_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_3", + "inputs": [ + "pow_4" + ], + "output": "mean_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_10", + "inputs": [ + "mean_3" + ], + "output": "add_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_5", + "inputs": [ + "add_10" + ], + "output": "rsqrt_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_15", + "inputs": [ + "_to_copy_18", + "rsqrt_5" + ], + "output": "mul_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_19", + "inputs": [ + "p_model_model_language_model_layers_1_input_layernorm_weight" + ], + "output": "_to_copy_19", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_11", + "inputs": [ + "_to_copy_19" + ], + "output": "add_11", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_16", + "inputs": [ + "mul_15", + "add_11" + ], + "output": "mul_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_2", + "inputs": [ + "mul_16" + ], + "output": "type_as_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_8", + "inputs": [ + "type_as_2", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_8", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_16", + "inputs": [ + "linear_8" + ], + "output": "transpose_16", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_9", + "inputs": [ + "type_as_2", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight" + ], + "output": "linear_9", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_7", + "inputs": [ + "linear_9" + ], + "output": "view_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_10", + "inputs": [ + "type_as_2", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight" + ], + "output": "linear_10", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_11", + "inputs": [ + "type_as_2", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight" + ], + "output": "linear_11", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_2", + "inputs": [ + "states_2", + "transpose_16" + ], + "output": "cat_2", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_18", + "inputs": [ + "cat_2" + ], + "output": "slice_18", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_2" + }, + { + "node": "squeeze_1", + "inputs": [ + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + ], + "output": "squeeze_1", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_6", + "inputs": [ + "squeeze_1" + ], + "output": "unsqueeze_6", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_1", + "inputs": [ + "cat_2", + "unsqueeze_6" + ], + "output": "conv1d_1", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_19", + "inputs": [ + "conv1d_1" + ], + "output": "slice_19", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_3", + "inputs": [ + "slice_19" + ], + "output": "silu_3", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_20", + "inputs": [ + "silu_3" + ], + "output": "slice_20", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_17", + "inputs": [ + "slice_20" + ], + "output": "transpose_17", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_21", + "inputs": [ + "transpose_17" + ], + "output": "slice_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_22", + "inputs": [ + "transpose_17" + ], + "output": "slice_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_23", + "inputs": [ + "transpose_17" + ], + "output": "slice_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_8", + "inputs": [ + "slice_21" + ], + "output": "view_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_9", + "inputs": [ + "slice_22" + ], + "output": "view_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_10", + "inputs": [ + "slice_23" + ], + "output": "view_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_1", + "inputs": [ + "linear_10" + ], + "output": "sigmoid_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_20", + "inputs": [ + "p_model_model_language_model_layers_1_linear_attn_a_log" + ], + "output": "_to_copy_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_1", + "inputs": [ + "_to_copy_20" + ], + "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": "_to_copy_21", + "inputs": [ + "linear_11" + ], + "output": "_to_copy_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_12", + "inputs": [ + "_to_copy_21", + "p_model_model_language_model_layers_1_linear_attn_dt_bias" + ], + "output": "add_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_8" + ], + "output": "mul_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_3", + "inputs": [ + "mul_18" + ], + "output": "sum_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_13", + "inputs": [ + "sum_3" + ], + "output": "add_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_6", + "inputs": [ + "add_13" + ], + "output": "rsqrt_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_19", + "inputs": [ + "view_8", + "rsqrt_6" + ], + "output": "mul_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_20", + "inputs": [ + "view_9" + ], + "output": "mul_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_4", + "inputs": [ + "mul_20" + ], + "output": "sum_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_14", + "inputs": [ + "sum_4" + ], + "output": "add_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_7", + "inputs": [ + "add_14" + ], + "output": "rsqrt_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_21", + "inputs": [ + "view_9", + "rsqrt_7" + ], + "output": "mul_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_22", + "inputs": [ + "mul_19" + ], + "output": "_to_copy_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_23", + "inputs": [ + "mul_21" + ], + "output": "_to_copy_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_24", + "inputs": [ + "view_10" + ], + "output": "_to_copy_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_25", + "inputs": [ + "sigmoid_1" + ], + "output": "_to_copy_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_1", + "inputs": [ + "_to_copy_22", + "_to_copy_23", + "_to_copy_24", + "mul_17", + "_to_copy_25", + "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": "clone_1", + "inputs": [ + "transpose_18" + ], + "output": "clone_1", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_26", + "inputs": [ + "clone_1" + ], + "output": "_to_copy_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_11", + "inputs": [ + "_to_copy_26" + ], + "output": "view_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_12", + "inputs": [ + "view_7" + ], + "output": "view_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_27", + "inputs": [ + "view_11" + ], + "output": "_to_copy_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_5", + "inputs": [ + "_to_copy_27" + ], + "output": "pow_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_4", + "inputs": [ + "pow_5" + ], + "output": "mean_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_15", + "inputs": [ + "mean_4" + ], + "output": "add_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_8", + "inputs": [ + "add_15" + ], + "output": "rsqrt_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_22", + "inputs": [ + "_to_copy_27", + "rsqrt_8" + ], + "output": "mul_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_28", + "inputs": [ + "mul_22" + ], + "output": "_to_copy_28", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_23", + "inputs": [ + "p_model_model_language_model_layers_1_linear_attn_norm_weight", + "_to_copy_28" + ], + "output": "mul_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_29", + "inputs": [ + "view_12" + ], + "output": "_to_copy_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_24", + "inputs": [ + "_to_copy_29", + "mul_23" + ], + "output": "mul_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_30", + "inputs": [ + "mul_24" + ], + "output": "_to_copy_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_13", + "inputs": [ + "_to_copy_30" + ], + "output": "view_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_12", + "inputs": [ + "view_13", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight" + ], + "output": "linear_12", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_16", + "inputs": [ + "add_9", + "linear_12" + ], + "output": "add_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_31", + "inputs": [ + "add_16" + ], + "output": "_to_copy_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_6", + "inputs": [ + "_to_copy_31" + ], + "output": "pow_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_5", + "inputs": [ + "pow_6" + ], + "output": "mean_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_17", + "inputs": [ + "mean_5" + ], + "output": "add_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_9", + "inputs": [ + "add_17" + ], + "output": "rsqrt_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_25", + "inputs": [ + "_to_copy_31", + "rsqrt_9" + ], + "output": "mul_25", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_32", + "inputs": [ + "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + ], + "output": "_to_copy_32", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_18", + "inputs": [ + "_to_copy_32" + ], + "output": "add_18", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_26", + "inputs": [ + "mul_25", + "add_18" + ], + "output": "mul_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_3", + "inputs": [ + "mul_26" + ], + "output": "type_as_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_13", + "inputs": [ + "type_as_3", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight" + ], + "output": "linear_13", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_14", + "inputs": [ + "type_as_3", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_27", + "inputs": [ + "linear_13", + "linear_14" + ], + "output": "mul_27", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_15", + "inputs": [ + "mul_27", + "p_model_model_language_model_layers_1_mlp_down_proj_weight" + ], + "output": "linear_15", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_19", + "inputs": [ + "add_16", + "linear_15" + ], + "output": "add_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_33", + "inputs": [ + "add_19" + ], + "output": "_to_copy_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_7", + "inputs": [ + "_to_copy_33" + ], + "output": "pow_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_6", + "inputs": [ + "pow_7" + ], + "output": "mean_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_20", + "inputs": [ + "mean_6" + ], + "output": "add_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_10", + "inputs": [ + "add_20" + ], + "output": "rsqrt_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_28", + "inputs": [ + "_to_copy_33", + "rsqrt_10" + ], + "output": "mul_28", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_34", + "inputs": [ + "p_model_model_language_model_layers_2_input_layernorm_weight" + ], + "output": "_to_copy_34", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_21", + "inputs": [ + "_to_copy_34" + ], + "output": "add_21", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_29", + "inputs": [ + "mul_28", + "add_21" + ], + "output": "mul_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_4", + "inputs": [ + "mul_29" + ], + "output": "type_as_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_16", + "inputs": [ + "type_as_4", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_16", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_19", + "inputs": [ + "linear_16" + ], + "output": "transpose_19", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_17", + "inputs": [ + "type_as_4", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight" + ], + "output": "linear_17", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_14", + "inputs": [ + "linear_17" + ], + "output": "view_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_18", + "inputs": [ + "type_as_4", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight" + ], + "output": "linear_18", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_19", + "inputs": [ + "type_as_4", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight" + ], + "output": "linear_19", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_3", + "inputs": [ + "states_4", + "transpose_19" + ], + "output": "cat_3", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_26", + "inputs": [ + "cat_3" + ], + "output": "slice_26", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_4" + }, + { + "node": "squeeze_2", + "inputs": [ + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + ], + "output": "squeeze_2", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_7", + "inputs": [ + "squeeze_2" + ], + "output": "unsqueeze_7", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_2", + "inputs": [ + "cat_3", + "unsqueeze_7" + ], + "output": "conv1d_2", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_27", + "inputs": [ + "conv1d_2" + ], + "output": "slice_27", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_6", + "inputs": [ + "slice_27" + ], + "output": "silu_6", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_28", + "inputs": [ + "silu_6" + ], + "output": "slice_28", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_20", + "inputs": [ + "slice_28" + ], + "output": "transpose_20", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_29", + "inputs": [ + "transpose_20" + ], + "output": "slice_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_30", + "inputs": [ + "transpose_20" + ], + "output": "slice_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_31", + "inputs": [ + "transpose_20" + ], + "output": "slice_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_15", + "inputs": [ + "slice_29" + ], + "output": "view_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_16", + "inputs": [ + "slice_30" + ], + "output": "view_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_17", + "inputs": [ + "slice_31" + ], + "output": "view_17", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_2", + "inputs": [ + "linear_18" + ], + "output": "sigmoid_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_35", + "inputs": [ + "p_model_model_language_model_layers_2_linear_attn_a_log" + ], + "output": "_to_copy_35", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_2", + "inputs": [ + "_to_copy_35" + ], + "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": "_to_copy_36", + "inputs": [ + "linear_19" + ], + "output": "_to_copy_36", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_22", + "inputs": [ + "_to_copy_36", + "p_model_model_language_model_layers_2_linear_attn_dt_bias" + ], + "output": "add_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_15" + ], + "output": "mul_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_5", + "inputs": [ + "mul_31" + ], + "output": "sum_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_23", + "inputs": [ + "sum_5" + ], + "output": "add_23", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_11", + "inputs": [ + "add_23" + ], + "output": "rsqrt_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_32", + "inputs": [ + "view_15", + "rsqrt_11" + ], + "output": "mul_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_33", + "inputs": [ + "view_16" + ], + "output": "mul_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_6", + "inputs": [ + "mul_33" + ], + "output": "sum_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_24", + "inputs": [ + "sum_6" + ], + "output": "add_24", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_12", + "inputs": [ + "add_24" + ], + "output": "rsqrt_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_34", + "inputs": [ + "view_16", + "rsqrt_12" + ], + "output": "mul_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_37", + "inputs": [ + "mul_32" + ], + "output": "_to_copy_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_38", + "inputs": [ + "mul_34" + ], + "output": "_to_copy_38", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_39", + "inputs": [ + "view_17" + ], + "output": "_to_copy_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_40", + "inputs": [ + "sigmoid_2" + ], + "output": "_to_copy_40", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_2", + "inputs": [ + "_to_copy_37", + "_to_copy_38", + "_to_copy_39", + "mul_30", + "_to_copy_40", + "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": "clone_2", + "inputs": [ + "transpose_21" + ], + "output": "clone_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_41", + "inputs": [ + "clone_2" + ], + "output": "_to_copy_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_18", + "inputs": [ + "_to_copy_41" + ], + "output": "view_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_19", + "inputs": [ + "view_14" + ], + "output": "view_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_42", + "inputs": [ + "view_18" + ], + "output": "_to_copy_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_8", + "inputs": [ + "_to_copy_42" + ], + "output": "pow_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_7", + "inputs": [ + "pow_8" + ], + "output": "mean_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_25", + "inputs": [ + "mean_7" + ], + "output": "add_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_13", + "inputs": [ + "add_25" + ], + "output": "rsqrt_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_35", + "inputs": [ + "_to_copy_42", + "rsqrt_13" + ], + "output": "mul_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_43", + "inputs": [ + "mul_35" + ], + "output": "_to_copy_43", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_36", + "inputs": [ + "p_model_model_language_model_layers_2_linear_attn_norm_weight", + "_to_copy_43" + ], + "output": "mul_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_44", + "inputs": [ + "view_19" + ], + "output": "_to_copy_44", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_37", + "inputs": [ + "_to_copy_44", + "mul_36" + ], + "output": "mul_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_45", + "inputs": [ + "mul_37" + ], + "output": "_to_copy_45", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_20", + "inputs": [ + "_to_copy_45" + ], + "output": "view_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_20", + "inputs": [ + "view_20", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight" + ], + "output": "linear_20", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_26", + "inputs": [ + "add_19", + "linear_20" + ], + "output": "add_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_46", + "inputs": [ + "add_26" + ], + "output": "_to_copy_46", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_9", + "inputs": [ + "_to_copy_46" + ], + "output": "pow_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_8", + "inputs": [ + "pow_9" + ], + "output": "mean_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_27", + "inputs": [ + "mean_8" + ], + "output": "add_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_14", + "inputs": [ + "add_27" + ], + "output": "rsqrt_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_38", + "inputs": [ + "_to_copy_46", + "rsqrt_14" + ], + "output": "mul_38", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_47", + "inputs": [ + "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + ], + "output": "_to_copy_47", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_28", + "inputs": [ + "_to_copy_47" + ], + "output": "add_28", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_39", + "inputs": [ + "mul_38", + "add_28" + ], + "output": "mul_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_5", + "inputs": [ + "mul_39" + ], + "output": "type_as_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_21", + "inputs": [ + "type_as_5", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight" + ], + "output": "linear_21", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_22", + "inputs": [ + "type_as_5", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_40", + "inputs": [ + "linear_21", + "linear_22" + ], + "output": "mul_40", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_23", + "inputs": [ + "mul_40", + "p_model_model_language_model_layers_2_mlp_down_proj_weight" + ], + "output": "linear_23", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_29", + "inputs": [ + "add_26", + "linear_23" + ], + "output": "add_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_48", + "inputs": [ + "add_29" + ], + "output": "_to_copy_48", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_10", + "inputs": [ + "_to_copy_48" + ], + "output": "pow_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_9", + "inputs": [ + "pow_10" + ], + "output": "mean_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_30", + "inputs": [ + "mean_9" + ], + "output": "add_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_15", + "inputs": [ + "add_30" + ], + "output": "rsqrt_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_41", + "inputs": [ + "_to_copy_48", + "rsqrt_15" + ], + "output": "mul_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_49", + "inputs": [ + "p_model_model_language_model_layers_3_input_layernorm_weight" + ], + "output": "_to_copy_49", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_31", + "inputs": [ + "_to_copy_49" + ], + "output": "add_31", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_42", + "inputs": [ + "mul_41", + "add_31" + ], + "output": "mul_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_6", + "inputs": [ + "mul_42" + ], + "output": "type_as_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_24", + "inputs": [ + "type_as_6", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight" + ], + "output": "linear_24", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl" + }, + { + "node": "view_21", + "inputs": [ + "linear_24" + ], + "output": "view_21", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl" + }, + { + "node": "slice_34", + "inputs": [ + "view_21" + ], + "output": "slice_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c956fa9d3342793cbd859a6e47baea9cfafaac162fc2accf4b8f4728d0f604.wgsl" + }, + { + "node": "slice_35", + "inputs": [ + "view_21" + ], + "output": "slice_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl" + }, + { + "node": "clone_3", + "inputs": [ + "slice_35" + ], + "output": "clone_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_unsafe_view", + "inputs": [ + "clone_3" + ], + "output": "_unsafe_view", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_22", + "inputs": [ + "slice_34" + ], + "output": "view_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_50", + "inputs": [ + "view_22" + ], + "output": "_to_copy_50", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_11", + "inputs": [ + "_to_copy_50" + ], + "output": "pow_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_10", + "inputs": [ + "pow_11" + ], + "output": "mean_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl" + }, + { + "node": "add_32", + "inputs": [ + "mean_10" + ], + "output": "add_32", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl" + }, + { + "node": "rsqrt_16", + "inputs": [ + "add_32" + ], + "output": "rsqrt_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl" + }, + { + "node": "mul_43", + "inputs": [ + "_to_copy_50", + "rsqrt_16" + ], + "output": "mul_43", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl" + }, + { + "node": "_to_copy_51", + "inputs": [ + "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + ], + "output": "_to_copy_51", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_33", + "inputs": [ + "_to_copy_51" + ], + "output": "add_33", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_44", + "inputs": [ + "mul_43", + "add_33" + ], + "output": "mul_44", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl" + }, + { + "node": "type_as_7", + "inputs": [ + "mul_44" + ], + "output": "type_as_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "transpose_22", + "inputs": [ + "type_as_7" + ], + "output": "transpose_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa1c36207dbf73b89a207971c82a8ed3c252e4b5057c44d46f9646ecce99da76.wgsl" + }, + { + "node": "linear_25", + "inputs": [ + "type_as_6", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight" + ], + "output": "linear_25", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_23", + "inputs": [ + "linear_25" + ], + "output": "view_23", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "_to_copy_52", + "inputs": [ + "view_23" + ], + "output": "_to_copy_52", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "pow_12", + "inputs": [ + "_to_copy_52" + ], + "output": "pow_12", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl" + }, + { + "node": "mean_11", + "inputs": [ + "pow_12" + ], + "output": "mean_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl" + }, + { + "node": "add_34", + "inputs": [ + "mean_11" + ], + "output": "add_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl" + }, + { + "node": "rsqrt_17", + "inputs": [ + "add_34" + ], + "output": "rsqrt_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl" + }, + { + "node": "mul_45", + "inputs": [ + "_to_copy_52", + "rsqrt_17" + ], + "output": "mul_45", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl" + }, + { + "node": "_to_copy_53", + "inputs": [ + "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + ], + "output": "_to_copy_53", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_35", + "inputs": [ + "_to_copy_53" + ], + "output": "add_35", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_46", + "inputs": [ + "mul_45", + "add_35" + ], + "output": "mul_46", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl" + }, + { + "node": "type_as_8", + "inputs": [ + "mul_46" + ], + "output": "type_as_8", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f8a4395bd0356791917d7ef0469f9b7d983d33d4b7ad044d66852ead8a86971b.wgsl" + }, + { + "node": "transpose_23", + "inputs": [ + "type_as_8" + ], + "output": "transpose_23", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "linear_26", + "inputs": [ + "type_as_6", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight" + ], + "output": "linear_26", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_24", + "inputs": [ + "linear_26" + ], + "output": "view_24", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "transpose_24", + "inputs": [ + "view_24" + ], + "output": "transpose_24", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "unsqueeze_8", + "inputs": [ + "_to_copy_1" + ], + "output": "unsqueeze_8", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "unsqueeze_9", + "inputs": [ + "_to_copy_2" + ], + "output": "unsqueeze_9", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "slice_36", + "inputs": [ + "transpose_22" + ], + "output": "slice_36", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl" + }, + { + "node": "slice_37", + "inputs": [ + "transpose_22" + ], + "output": "slice_37", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl" + }, + { + "node": "slice_38", + "inputs": [ + "transpose_23" + ], + "output": "slice_38", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e697b1bc29cd0d08bba0c6a455a7e4b02a3c2e2535304faff2c04b6976617c7c.wgsl" + }, + { + "node": "slice_39", + "inputs": [ + "transpose_23" + ], + "output": "slice_39", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl" + }, + { + "node": "mul_47", + "inputs": [ + "slice_36", + "unsqueeze_8" + ], + "output": "mul_47", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "slice_40", + "inputs": [ + "slice_36" + ], + "output": "slice_40", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl" + }, + { + "node": "slice_41", + "inputs": [ + "slice_36" + ], + "output": "slice_41", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl" + }, + { + "node": "neg_3", + "inputs": [ + "slice_41" + ], + "output": "neg_3", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl" + }, + { + "node": "cat_4", + "inputs": [ + "neg_3", + "slice_40" + ], + "output": "cat_4", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl" + }, + { + "node": "mul_48", + "inputs": [ + "cat_4", + "unsqueeze_9" + ], + "output": "mul_48", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "add_36", + "inputs": [ + "mul_47", + "mul_48" + ], + "output": "add_36", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl" + }, + { + "node": "mul_49", + "inputs": [ + "slice_38", + "unsqueeze_8" + ], + "output": "mul_49", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "slice_42", + "inputs": [ + "slice_38" + ], + "output": "slice_42", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/fd2084f233c932d1160371a9a522dbc8abdf416285eef54e02e18e4738d12675.wgsl" + }, + { + "node": "slice_43", + "inputs": [ + "slice_38" + ], + "output": "slice_43", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl" + }, + { + "node": "neg_4", + "inputs": [ + "slice_43" + ], + "output": "neg_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl" + }, + { + "node": "cat_5", + "inputs": [ + "neg_4", + "slice_42" + ], + "output": "cat_5", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl" + }, + { + "node": "mul_50", + "inputs": [ + "cat_5", + "unsqueeze_9" + ], + "output": "mul_50", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "add_37", + "inputs": [ + "mul_49", + "mul_50" + ], + "output": "add_37", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl" + }, + { + "node": "cat_6", + "inputs": [ + "add_36", + "slice_37" + ], + "output": "cat_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl" + }, + { + "node": "cat_7", + "inputs": [ + "add_37", + "slice_39" + ], + "output": "cat_7", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl" + }, + { + "node": "cache_update", + "inputs": [ + "cat_7", + "position_ids" + ], + "output": "cache_update", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_6", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "cache_update_1", + "inputs": [ + "transpose_24", + "position_ids" + ], + "output": "cache_update_1", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_7", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "unsqueeze_10", + "inputs": [ + "cache_update" + ], + "output": "unsqueeze_10", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_2", + "inputs": [ + "unsqueeze_10" + ], + "output": "expand_2", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_4", + "inputs": [ + "expand_2" + ], + "output": "clone_4", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_1", + "inputs": [ + "clone_4" + ], + "output": "_unsafe_view_1", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "unsqueeze_11", + "inputs": [ + "cache_update_1" + ], + "output": "unsqueeze_11", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_3", + "inputs": [ + "unsqueeze_11" + ], + "output": "expand_3", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_5", + "inputs": [ + "expand_3" + ], + "output": "clone_5", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_2", + "inputs": [ + "clone_5" + ], + "output": "_unsafe_view_2", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "transpose_25", + "inputs": [ + "_unsafe_view_1" + ], + "output": "transpose_25", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl" + }, + { + "node": "matmul_1", + "inputs": [ + "cat_6", + "transpose_25" + ], + "output": "matmul_1", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl" + }, + { + "node": "mul_51", + "inputs": [ + "matmul_1" + ], + "output": "mul_51", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl" + }, + { + "node": "add_38", + "inputs": [ + "mul_51", + "attention_mask" + ], + "output": "add_38", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl" + }, + { + "node": "softmax", + "inputs": [ + "add_38" + ], + "output": "softmax", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f40727952ccdc71060abbdab6a141b0497a38b727d2f5b943fe826ce90497bd1.wgsl" + }, + { + "node": "_to_copy_54", + "inputs": [ + "softmax" + ], + "output": "_to_copy_54", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl" + }, + { + "node": "matmul_2", + "inputs": [ + "_to_copy_54", + "_unsafe_view_2" + ], + "output": "matmul_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl" + }, + { + "node": "transpose_26", + "inputs": [ + "matmul_2" + ], + "output": "transpose_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl" + }, + { + "node": "clone_6", + "inputs": [ + "transpose_26" + ], + "output": "clone_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_25", + "inputs": [ + "clone_6" + ], + "output": "view_25", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_3", + "inputs": [ + "_unsafe_view" + ], + "output": "sigmoid_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl" + }, + { + "node": "mul_52", + "inputs": [ + "view_25", + "sigmoid_3" + ], + "output": "mul_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/efd4310107ed3db32f1792f1360971ff7ce4234ab64fda03dcaaa9bff52bb912.wgsl" + }, + { + "node": "linear_27", + "inputs": [ + "mul_52", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight" + ], + "output": "linear_27", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_39", + "inputs": [ + "add_29", + "linear_27" + ], + "output": "add_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_55", + "inputs": [ + "add_39" + ], + "output": "_to_copy_55", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_13", + "inputs": [ + "_to_copy_55" + ], + "output": "pow_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_12", + "inputs": [ + "pow_13" + ], + "output": "mean_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_40", + "inputs": [ + "mean_12" + ], + "output": "add_40", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_18", + "inputs": [ + "add_40" + ], + "output": "rsqrt_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_53", + "inputs": [ + "_to_copy_55", + "rsqrt_18" + ], + "output": "mul_53", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_56", + "inputs": [ + "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + ], + "output": "_to_copy_56", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_41", + "inputs": [ + "_to_copy_56" + ], + "output": "add_41", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_54", + "inputs": [ + "mul_53", + "add_41" + ], + "output": "mul_54", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_9", + "inputs": [ + "mul_54" + ], + "output": "type_as_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_28", + "inputs": [ + "type_as_9", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight" + ], + "output": "linear_28", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_29", + "inputs": [ + "type_as_9", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_55", + "inputs": [ + "linear_28", + "linear_29" + ], + "output": "mul_55", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_30", + "inputs": [ + "mul_55", + "p_model_model_language_model_layers_3_mlp_down_proj_weight" + ], + "output": "linear_30", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_42", + "inputs": [ + "add_39", + "linear_30" + ], + "output": "add_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_57", + "inputs": [ + "add_42" + ], + "output": "_to_copy_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_14", + "inputs": [ + "_to_copy_57" + ], + "output": "pow_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_13", + "inputs": [ + "pow_14" + ], + "output": "mean_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_43", + "inputs": [ + "mean_13" + ], + "output": "add_43", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_19", + "inputs": [ + "add_43" + ], + "output": "rsqrt_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_56", + "inputs": [ + "_to_copy_57", + "rsqrt_19" + ], + "output": "mul_56", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_58", + "inputs": [ + "p_model_model_language_model_layers_4_input_layernorm_weight" + ], + "output": "_to_copy_58", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_44", + "inputs": [ + "_to_copy_58" + ], + "output": "add_44", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_57", + "inputs": [ + "mul_56", + "add_44" + ], + "output": "mul_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_10", + "inputs": [ + "mul_57" + ], + "output": "type_as_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_31", + "inputs": [ + "type_as_10", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_31", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_27", + "inputs": [ + "linear_31" + ], + "output": "transpose_27", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_32", + "inputs": [ + "type_as_10", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight" + ], + "output": "linear_32", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_26", + "inputs": [ + "linear_32" + ], + "output": "view_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_33", + "inputs": [ + "type_as_10", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight" + ], + "output": "linear_33", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_34", + "inputs": [ + "type_as_10", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight" + ], + "output": "linear_34", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_8", + "inputs": [ + "states_8", + "transpose_27" + ], + "output": "cat_8", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_44", + "inputs": [ + "cat_8" + ], + "output": "slice_44", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_8" + }, + { + "node": "squeeze_3", + "inputs": [ + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + ], + "output": "squeeze_3", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_12", + "inputs": [ + "squeeze_3" + ], + "output": "unsqueeze_12", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_3", + "inputs": [ + "cat_8", + "unsqueeze_12" + ], + "output": "conv1d_3", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_45", + "inputs": [ + "conv1d_3" + ], + "output": "slice_45", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_10", + "inputs": [ + "slice_45" + ], + "output": "silu_10", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_46", + "inputs": [ + "silu_10" + ], + "output": "slice_46", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_28", + "inputs": [ + "slice_46" + ], + "output": "transpose_28", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_47", + "inputs": [ + "transpose_28" + ], + "output": "slice_47", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_48", + "inputs": [ + "transpose_28" + ], + "output": "slice_48", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_49", + "inputs": [ + "transpose_28" + ], + "output": "slice_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_27", + "inputs": [ + "slice_47" + ], + "output": "view_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_28", + "inputs": [ + "slice_48" + ], + "output": "view_28", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_29", + "inputs": [ + "slice_49" + ], + "output": "view_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_4", + "inputs": [ + "linear_33" + ], + "output": "sigmoid_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_59", + "inputs": [ + "p_model_model_language_model_layers_4_linear_attn_a_log" + ], + "output": "_to_copy_59", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_3", + "inputs": [ + "_to_copy_59" + ], + "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": "_to_copy_60", + "inputs": [ + "linear_34" + ], + "output": "_to_copy_60", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_45", + "inputs": [ + "_to_copy_60", + "p_model_model_language_model_layers_4_linear_attn_dt_bias" + ], + "output": "add_45", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_27" + ], + "output": "mul_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_7", + "inputs": [ + "mul_59" + ], + "output": "sum_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_46", + "inputs": [ + "sum_7" + ], + "output": "add_46", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_20", + "inputs": [ + "add_46" + ], + "output": "rsqrt_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_60", + "inputs": [ + "view_27", + "rsqrt_20" + ], + "output": "mul_60", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_61", + "inputs": [ + "view_28" + ], + "output": "mul_61", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_8", + "inputs": [ + "mul_61" + ], + "output": "sum_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_47", + "inputs": [ + "sum_8" + ], + "output": "add_47", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_21", + "inputs": [ + "add_47" + ], + "output": "rsqrt_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_62", + "inputs": [ + "view_28", + "rsqrt_21" + ], + "output": "mul_62", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_61", + "inputs": [ + "mul_60" + ], + "output": "_to_copy_61", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_62", + "inputs": [ + "mul_62" + ], + "output": "_to_copy_62", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_63", + "inputs": [ + "view_29" + ], + "output": "_to_copy_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_64", + "inputs": [ + "sigmoid_4" + ], + "output": "_to_copy_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_3", + "inputs": [ + "_to_copy_61", + "_to_copy_62", + "_to_copy_63", + "mul_58", + "_to_copy_64", + "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": "clone_7", + "inputs": [ + "transpose_29" + ], + "output": "clone_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_65", + "inputs": [ + "clone_7" + ], + "output": "_to_copy_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_30", + "inputs": [ + "_to_copy_65" + ], + "output": "view_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_31", + "inputs": [ + "view_26" + ], + "output": "view_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_66", + "inputs": [ + "view_30" + ], + "output": "_to_copy_66", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_15", + "inputs": [ + "_to_copy_66" + ], + "output": "pow_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_14", + "inputs": [ + "pow_15" + ], + "output": "mean_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_48", + "inputs": [ + "mean_14" + ], + "output": "add_48", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_22", + "inputs": [ + "add_48" + ], + "output": "rsqrt_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_63", + "inputs": [ + "_to_copy_66", + "rsqrt_22" + ], + "output": "mul_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_67", + "inputs": [ + "mul_63" + ], + "output": "_to_copy_67", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_64", + "inputs": [ + "p_model_model_language_model_layers_4_linear_attn_norm_weight", + "_to_copy_67" + ], + "output": "mul_64", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_68", + "inputs": [ + "view_31" + ], + "output": "_to_copy_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_65", + "inputs": [ + "_to_copy_68", + "mul_64" + ], + "output": "mul_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_69", + "inputs": [ + "mul_65" + ], + "output": "_to_copy_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_32", + "inputs": [ + "_to_copy_69" + ], + "output": "view_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_35", + "inputs": [ + "view_32", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight" + ], + "output": "linear_35", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_49", + "inputs": [ + "add_42", + "linear_35" + ], + "output": "add_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_70", + "inputs": [ + "add_49" + ], + "output": "_to_copy_70", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_16", + "inputs": [ + "_to_copy_70" + ], + "output": "pow_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_15", + "inputs": [ + "pow_16" + ], + "output": "mean_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_50", + "inputs": [ + "mean_15" + ], + "output": "add_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_23", + "inputs": [ + "add_50" + ], + "output": "rsqrt_23", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_66", + "inputs": [ + "_to_copy_70", + "rsqrt_23" + ], + "output": "mul_66", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_71", + "inputs": [ + "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + ], + "output": "_to_copy_71", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_51", + "inputs": [ + "_to_copy_71" + ], + "output": "add_51", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_67", + "inputs": [ + "mul_66", + "add_51" + ], + "output": "mul_67", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_11", + "inputs": [ + "mul_67" + ], + "output": "type_as_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_36", + "inputs": [ + "type_as_11", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight" + ], + "output": "linear_36", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_37", + "inputs": [ + "type_as_11", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_68", + "inputs": [ + "linear_36", + "linear_37" + ], + "output": "mul_68", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_38", + "inputs": [ + "mul_68", + "p_model_model_language_model_layers_4_mlp_down_proj_weight" + ], + "output": "linear_38", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_52", + "inputs": [ + "add_49", + "linear_38" + ], + "output": "add_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_72", + "inputs": [ + "add_52" + ], + "output": "_to_copy_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_17", + "inputs": [ + "_to_copy_72" + ], + "output": "pow_17", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_16", + "inputs": [ + "pow_17" + ], + "output": "mean_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_53", + "inputs": [ + "mean_16" + ], + "output": "add_53", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_24", + "inputs": [ + "add_53" + ], + "output": "rsqrt_24", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_69", + "inputs": [ + "_to_copy_72", + "rsqrt_24" + ], + "output": "mul_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_73", + "inputs": [ + "p_model_model_language_model_layers_5_input_layernorm_weight" + ], + "output": "_to_copy_73", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_54", + "inputs": [ + "_to_copy_73" + ], + "output": "add_54", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_70", + "inputs": [ + "mul_69", + "add_54" + ], + "output": "mul_70", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_12", + "inputs": [ + "mul_70" + ], + "output": "type_as_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_39", + "inputs": [ + "type_as_12", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_39", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_30", + "inputs": [ + "linear_39" + ], + "output": "transpose_30", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_40", + "inputs": [ + "type_as_12", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight" + ], + "output": "linear_40", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_33", + "inputs": [ + "linear_40" + ], + "output": "view_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_41", + "inputs": [ + "type_as_12", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight" + ], + "output": "linear_41", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_42", + "inputs": [ + "type_as_12", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight" + ], + "output": "linear_42", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_9", + "inputs": [ + "states_10", + "transpose_30" + ], + "output": "cat_9", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_52", + "inputs": [ + "cat_9" + ], + "output": "slice_52", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_10" + }, + { + "node": "squeeze_4", + "inputs": [ + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + ], + "output": "squeeze_4", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_13", + "inputs": [ + "squeeze_4" + ], + "output": "unsqueeze_13", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_4", + "inputs": [ + "cat_9", + "unsqueeze_13" + ], + "output": "conv1d_4", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_53", + "inputs": [ + "conv1d_4" + ], + "output": "slice_53", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_13", + "inputs": [ + "slice_53" + ], + "output": "silu_13", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_54", + "inputs": [ + "silu_13" + ], + "output": "slice_54", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_31", + "inputs": [ + "slice_54" + ], + "output": "transpose_31", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_55", + "inputs": [ + "transpose_31" + ], + "output": "slice_55", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_56", + "inputs": [ + "transpose_31" + ], + "output": "slice_56", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_57", + "inputs": [ + "transpose_31" + ], + "output": "slice_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_34", + "inputs": [ + "slice_55" + ], + "output": "view_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_35", + "inputs": [ + "slice_56" + ], + "output": "view_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_36", + "inputs": [ + "slice_57" + ], + "output": "view_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_5", + "inputs": [ + "linear_41" + ], + "output": "sigmoid_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_74", + "inputs": [ + "p_model_model_language_model_layers_5_linear_attn_a_log" + ], + "output": "_to_copy_74", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_4", + "inputs": [ + "_to_copy_74" + ], + "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": "_to_copy_75", + "inputs": [ + "linear_42" + ], + "output": "_to_copy_75", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_55", + "inputs": [ + "_to_copy_75", + "p_model_model_language_model_layers_5_linear_attn_dt_bias" + ], + "output": "add_55", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_34" + ], + "output": "mul_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_9", + "inputs": [ + "mul_72" + ], + "output": "sum_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_56", + "inputs": [ + "sum_9" + ], + "output": "add_56", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_25", + "inputs": [ + "add_56" + ], + "output": "rsqrt_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_73", + "inputs": [ + "view_34", + "rsqrt_25" + ], + "output": "mul_73", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_74", + "inputs": [ + "view_35" + ], + "output": "mul_74", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_10", + "inputs": [ + "mul_74" + ], + "output": "sum_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_57", + "inputs": [ + "sum_10" + ], + "output": "add_57", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_26", + "inputs": [ + "add_57" + ], + "output": "rsqrt_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_75", + "inputs": [ + "view_35", + "rsqrt_26" + ], + "output": "mul_75", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_76", + "inputs": [ + "mul_73" + ], + "output": "_to_copy_76", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_77", + "inputs": [ + "mul_75" + ], + "output": "_to_copy_77", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_78", + "inputs": [ + "view_36" + ], + "output": "_to_copy_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_79", + "inputs": [ + "sigmoid_5" + ], + "output": "_to_copy_79", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_4", + "inputs": [ + "_to_copy_76", + "_to_copy_77", + "_to_copy_78", + "mul_71", + "_to_copy_79", + "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": "clone_8", + "inputs": [ + "transpose_32" + ], + "output": "clone_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_80", + "inputs": [ + "clone_8" + ], + "output": "_to_copy_80", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_37", + "inputs": [ + "_to_copy_80" + ], + "output": "view_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_38", + "inputs": [ + "view_33" + ], + "output": "view_38", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_81", + "inputs": [ + "view_37" + ], + "output": "_to_copy_81", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_18", + "inputs": [ + "_to_copy_81" + ], + "output": "pow_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_17", + "inputs": [ + "pow_18" + ], + "output": "mean_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_58", + "inputs": [ + "mean_17" + ], + "output": "add_58", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_27", + "inputs": [ + "add_58" + ], + "output": "rsqrt_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_76", + "inputs": [ + "_to_copy_81", + "rsqrt_27" + ], + "output": "mul_76", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_82", + "inputs": [ + "mul_76" + ], + "output": "_to_copy_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_77", + "inputs": [ + "p_model_model_language_model_layers_5_linear_attn_norm_weight", + "_to_copy_82" + ], + "output": "mul_77", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_83", + "inputs": [ + "view_38" + ], + "output": "_to_copy_83", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_78", + "inputs": [ + "_to_copy_83", + "mul_77" + ], + "output": "mul_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_84", + "inputs": [ + "mul_78" + ], + "output": "_to_copy_84", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_39", + "inputs": [ + "_to_copy_84" + ], + "output": "view_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_43", + "inputs": [ + "view_39", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight" + ], + "output": "linear_43", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_59", + "inputs": [ + "add_52", + "linear_43" + ], + "output": "add_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_85", + "inputs": [ + "add_59" + ], + "output": "_to_copy_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_19", + "inputs": [ + "_to_copy_85" + ], + "output": "pow_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_18", + "inputs": [ + "pow_19" + ], + "output": "mean_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_60", + "inputs": [ + "mean_18" + ], + "output": "add_60", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_28", + "inputs": [ + "add_60" + ], + "output": "rsqrt_28", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_79", + "inputs": [ + "_to_copy_85", + "rsqrt_28" + ], + "output": "mul_79", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_86", + "inputs": [ + "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + ], + "output": "_to_copy_86", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_61", + "inputs": [ + "_to_copy_86" + ], + "output": "add_61", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_80", + "inputs": [ + "mul_79", + "add_61" + ], + "output": "mul_80", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_13", + "inputs": [ + "mul_80" + ], + "output": "type_as_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_44", + "inputs": [ + "type_as_13", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight" + ], + "output": "linear_44", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_45", + "inputs": [ + "type_as_13", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_81", + "inputs": [ + "linear_44", + "linear_45" + ], + "output": "mul_81", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_46", + "inputs": [ + "mul_81", + "p_model_model_language_model_layers_5_mlp_down_proj_weight" + ], + "output": "linear_46", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_62", + "inputs": [ + "add_59", + "linear_46" + ], + "output": "add_62", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_87", + "inputs": [ + "add_62" + ], + "output": "_to_copy_87", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_20", + "inputs": [ + "_to_copy_87" + ], + "output": "pow_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_19", + "inputs": [ + "pow_20" + ], + "output": "mean_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_63", + "inputs": [ + "mean_19" + ], + "output": "add_63", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_29", + "inputs": [ + "add_63" + ], + "output": "rsqrt_29", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_82", + "inputs": [ + "_to_copy_87", + "rsqrt_29" + ], + "output": "mul_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_88", + "inputs": [ + "p_model_model_language_model_layers_6_input_layernorm_weight" + ], + "output": "_to_copy_88", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_64", + "inputs": [ + "_to_copy_88" + ], + "output": "add_64", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_83", + "inputs": [ + "mul_82", + "add_64" + ], + "output": "mul_83", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_14", + "inputs": [ + "mul_83" + ], + "output": "type_as_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_47", + "inputs": [ + "type_as_14", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_47", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_33", + "inputs": [ + "linear_47" + ], + "output": "transpose_33", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_48", + "inputs": [ + "type_as_14", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight" + ], + "output": "linear_48", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_40", + "inputs": [ + "linear_48" + ], + "output": "view_40", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_49", + "inputs": [ + "type_as_14", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight" + ], + "output": "linear_49", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_50", + "inputs": [ + "type_as_14", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight" + ], + "output": "linear_50", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_10", + "inputs": [ + "states_12", + "transpose_33" + ], + "output": "cat_10", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_60", + "inputs": [ + "cat_10" + ], + "output": "slice_60", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_12" + }, + { + "node": "squeeze_5", + "inputs": [ + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + ], + "output": "squeeze_5", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_14", + "inputs": [ + "squeeze_5" + ], + "output": "unsqueeze_14", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_5", + "inputs": [ + "cat_10", + "unsqueeze_14" + ], + "output": "conv1d_5", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_61", + "inputs": [ + "conv1d_5" + ], + "output": "slice_61", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_16", + "inputs": [ + "slice_61" + ], + "output": "silu_16", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_62", + "inputs": [ + "silu_16" + ], + "output": "slice_62", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_34", + "inputs": [ + "slice_62" + ], + "output": "transpose_34", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_63", + "inputs": [ + "transpose_34" + ], + "output": "slice_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_64", + "inputs": [ + "transpose_34" + ], + "output": "slice_64", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_65", + "inputs": [ + "transpose_34" + ], + "output": "slice_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_41", + "inputs": [ + "slice_63" + ], + "output": "view_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_42", + "inputs": [ + "slice_64" + ], + "output": "view_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_43", + "inputs": [ + "slice_65" + ], + "output": "view_43", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_6", + "inputs": [ + "linear_49" + ], + "output": "sigmoid_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_89", + "inputs": [ + "p_model_model_language_model_layers_6_linear_attn_a_log" + ], + "output": "_to_copy_89", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_5", + "inputs": [ + "_to_copy_89" + ], + "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": "_to_copy_90", + "inputs": [ + "linear_50" + ], + "output": "_to_copy_90", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_65", + "inputs": [ + "_to_copy_90", + "p_model_model_language_model_layers_6_linear_attn_dt_bias" + ], + "output": "add_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_41" + ], + "output": "mul_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_11", + "inputs": [ + "mul_85" + ], + "output": "sum_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_66", + "inputs": [ + "sum_11" + ], + "output": "add_66", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_30", + "inputs": [ + "add_66" + ], + "output": "rsqrt_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_86", + "inputs": [ + "view_41", + "rsqrt_30" + ], + "output": "mul_86", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_87", + "inputs": [ + "view_42" + ], + "output": "mul_87", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_12", + "inputs": [ + "mul_87" + ], + "output": "sum_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_67", + "inputs": [ + "sum_12" + ], + "output": "add_67", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_31", + "inputs": [ + "add_67" + ], + "output": "rsqrt_31", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_88", + "inputs": [ + "view_42", + "rsqrt_31" + ], + "output": "mul_88", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_91", + "inputs": [ + "mul_86" + ], + "output": "_to_copy_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_92", + "inputs": [ + "mul_88" + ], + "output": "_to_copy_92", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_93", + "inputs": [ + "view_43" + ], + "output": "_to_copy_93", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_94", + "inputs": [ + "sigmoid_6" + ], + "output": "_to_copy_94", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_5", + "inputs": [ + "_to_copy_91", + "_to_copy_92", + "_to_copy_93", + "mul_84", + "_to_copy_94", + "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": "clone_9", + "inputs": [ + "transpose_35" + ], + "output": "clone_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_95", + "inputs": [ + "clone_9" + ], + "output": "_to_copy_95", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_44", + "inputs": [ + "_to_copy_95" + ], + "output": "view_44", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_45", + "inputs": [ + "view_40" + ], + "output": "view_45", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_96", + "inputs": [ + "view_44" + ], + "output": "_to_copy_96", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_21", + "inputs": [ + "_to_copy_96" + ], + "output": "pow_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_20", + "inputs": [ + "pow_21" + ], + "output": "mean_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_68", + "inputs": [ + "mean_20" + ], + "output": "add_68", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_32", + "inputs": [ + "add_68" + ], + "output": "rsqrt_32", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_89", + "inputs": [ + "_to_copy_96", + "rsqrt_32" + ], + "output": "mul_89", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_97", + "inputs": [ + "mul_89" + ], + "output": "_to_copy_97", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_90", + "inputs": [ + "p_model_model_language_model_layers_6_linear_attn_norm_weight", + "_to_copy_97" + ], + "output": "mul_90", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_98", + "inputs": [ + "view_45" + ], + "output": "_to_copy_98", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_91", + "inputs": [ + "_to_copy_98", + "mul_90" + ], + "output": "mul_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_99", + "inputs": [ + "mul_91" + ], + "output": "_to_copy_99", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_46", + "inputs": [ + "_to_copy_99" + ], + "output": "view_46", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_51", + "inputs": [ + "view_46", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight" + ], + "output": "linear_51", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_69", + "inputs": [ + "add_62", + "linear_51" + ], + "output": "add_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_100", + "inputs": [ + "add_69" + ], + "output": "_to_copy_100", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_22", + "inputs": [ + "_to_copy_100" + ], + "output": "pow_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_21", + "inputs": [ + "pow_22" + ], + "output": "mean_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_70", + "inputs": [ + "mean_21" + ], + "output": "add_70", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_33", + "inputs": [ + "add_70" + ], + "output": "rsqrt_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_92", + "inputs": [ + "_to_copy_100", + "rsqrt_33" + ], + "output": "mul_92", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_101", + "inputs": [ + "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + ], + "output": "_to_copy_101", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_71", + "inputs": [ + "_to_copy_101" + ], + "output": "add_71", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_93", + "inputs": [ + "mul_92", + "add_71" + ], + "output": "mul_93", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_15", + "inputs": [ + "mul_93" + ], + "output": "type_as_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_52", + "inputs": [ + "type_as_15", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight" + ], + "output": "linear_52", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_53", + "inputs": [ + "type_as_15", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_94", + "inputs": [ + "linear_52", + "linear_53" + ], + "output": "mul_94", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_54", + "inputs": [ + "mul_94", + "p_model_model_language_model_layers_6_mlp_down_proj_weight" + ], + "output": "linear_54", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_72", + "inputs": [ + "add_69", + "linear_54" + ], + "output": "add_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_102", + "inputs": [ + "add_72" + ], + "output": "_to_copy_102", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_23", + "inputs": [ + "_to_copy_102" + ], + "output": "pow_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_22", + "inputs": [ + "pow_23" + ], + "output": "mean_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_73", + "inputs": [ + "mean_22" + ], + "output": "add_73", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_34", + "inputs": [ + "add_73" + ], + "output": "rsqrt_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_95", + "inputs": [ + "_to_copy_102", + "rsqrt_34" + ], + "output": "mul_95", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_103", + "inputs": [ + "p_model_model_language_model_layers_7_input_layernorm_weight" + ], + "output": "_to_copy_103", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_74", + "inputs": [ + "_to_copy_103" + ], + "output": "add_74", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_96", + "inputs": [ + "mul_95", + "add_74" + ], + "output": "mul_96", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_16", + "inputs": [ + "mul_96" + ], + "output": "type_as_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_55", + "inputs": [ + "type_as_16", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight" + ], + "output": "linear_55", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl" + }, + { + "node": "view_47", + "inputs": [ + "linear_55" + ], + "output": "view_47", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl" + }, + { + "node": "slice_68", + "inputs": [ + "view_47" + ], + "output": "slice_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c956fa9d3342793cbd859a6e47baea9cfafaac162fc2accf4b8f4728d0f604.wgsl" + }, + { + "node": "slice_69", + "inputs": [ + "view_47" + ], + "output": "slice_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl" + }, + { + "node": "clone_10", + "inputs": [ + "slice_69" + ], + "output": "clone_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_unsafe_view_3", + "inputs": [ + "clone_10" + ], + "output": "_unsafe_view_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_48", + "inputs": [ + "slice_68" + ], + "output": "view_48", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_104", + "inputs": [ + "view_48" + ], + "output": "_to_copy_104", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_24", + "inputs": [ + "_to_copy_104" + ], + "output": "pow_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_23", + "inputs": [ + "pow_24" + ], + "output": "mean_23", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl" + }, + { + "node": "add_75", + "inputs": [ + "mean_23" + ], + "output": "add_75", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl" + }, + { + "node": "rsqrt_35", + "inputs": [ + "add_75" + ], + "output": "rsqrt_35", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl" + }, + { + "node": "mul_97", + "inputs": [ + "_to_copy_104", + "rsqrt_35" + ], + "output": "mul_97", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl" + }, + { + "node": "_to_copy_105", + "inputs": [ + "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + ], + "output": "_to_copy_105", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_76", + "inputs": [ + "_to_copy_105" + ], + "output": "add_76", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_98", + "inputs": [ + "mul_97", + "add_76" + ], + "output": "mul_98", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl" + }, + { + "node": "type_as_17", + "inputs": [ + "mul_98" + ], + "output": "type_as_17", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "transpose_36", + "inputs": [ + "type_as_17" + ], + "output": "transpose_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa1c36207dbf73b89a207971c82a8ed3c252e4b5057c44d46f9646ecce99da76.wgsl" + }, + { + "node": "linear_56", + "inputs": [ + "type_as_16", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight" + ], + "output": "linear_56", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_49", + "inputs": [ + "linear_56" + ], + "output": "view_49", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "_to_copy_106", + "inputs": [ + "view_49" + ], + "output": "_to_copy_106", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "pow_25", + "inputs": [ + "_to_copy_106" + ], + "output": "pow_25", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl" + }, + { + "node": "mean_24", + "inputs": [ + "pow_25" + ], + "output": "mean_24", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl" + }, + { + "node": "add_77", + "inputs": [ + "mean_24" + ], + "output": "add_77", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl" + }, + { + "node": "rsqrt_36", + "inputs": [ + "add_77" + ], + "output": "rsqrt_36", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl" + }, + { + "node": "mul_99", + "inputs": [ + "_to_copy_106", + "rsqrt_36" + ], + "output": "mul_99", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl" + }, + { + "node": "_to_copy_107", + "inputs": [ + "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + ], + "output": "_to_copy_107", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_78", + "inputs": [ + "_to_copy_107" + ], + "output": "add_78", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_100", + "inputs": [ + "mul_99", + "add_78" + ], + "output": "mul_100", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl" + }, + { + "node": "type_as_18", + "inputs": [ + "mul_100" + ], + "output": "type_as_18", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f8a4395bd0356791917d7ef0469f9b7d983d33d4b7ad044d66852ead8a86971b.wgsl" + }, + { + "node": "transpose_37", + "inputs": [ + "type_as_18" + ], + "output": "transpose_37", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "linear_57", + "inputs": [ + "type_as_16", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight" + ], + "output": "linear_57", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_50", + "inputs": [ + "linear_57" + ], + "output": "view_50", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "transpose_38", + "inputs": [ + "view_50" + ], + "output": "transpose_38", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "unsqueeze_15", + "inputs": [ + "_to_copy_1" + ], + "output": "unsqueeze_15", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "unsqueeze_16", + "inputs": [ + "_to_copy_2" + ], + "output": "unsqueeze_16", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "slice_70", + "inputs": [ + "transpose_36" + ], + "output": "slice_70", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl" + }, + { + "node": "slice_71", + "inputs": [ + "transpose_36" + ], + "output": "slice_71", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl" + }, + { + "node": "slice_72", + "inputs": [ + "transpose_37" + ], + "output": "slice_72", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e697b1bc29cd0d08bba0c6a455a7e4b02a3c2e2535304faff2c04b6976617c7c.wgsl" + }, + { + "node": "slice_73", + "inputs": [ + "transpose_37" + ], + "output": "slice_73", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl" + }, + { + "node": "mul_101", + "inputs": [ + "slice_70", + "unsqueeze_15" + ], + "output": "mul_101", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "slice_74", + "inputs": [ + "slice_70" + ], + "output": "slice_74", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl" + }, + { + "node": "slice_75", + "inputs": [ + "slice_70" + ], + "output": "slice_75", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl" + }, + { + "node": "neg_8", + "inputs": [ + "slice_75" + ], + "output": "neg_8", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl" + }, + { + "node": "cat_11", + "inputs": [ + "neg_8", + "slice_74" + ], + "output": "cat_11", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl" + }, + { + "node": "mul_102", + "inputs": [ + "cat_11", + "unsqueeze_16" + ], + "output": "mul_102", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "add_79", + "inputs": [ + "mul_101", + "mul_102" + ], + "output": "add_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl" + }, + { + "node": "mul_103", + "inputs": [ + "slice_72", + "unsqueeze_15" + ], + "output": "mul_103", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "slice_76", + "inputs": [ + "slice_72" + ], + "output": "slice_76", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/fd2084f233c932d1160371a9a522dbc8abdf416285eef54e02e18e4738d12675.wgsl" + }, + { + "node": "slice_77", + "inputs": [ + "slice_72" + ], + "output": "slice_77", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl" + }, + { + "node": "neg_9", + "inputs": [ + "slice_77" + ], + "output": "neg_9", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl" + }, + { + "node": "cat_12", + "inputs": [ + "neg_9", + "slice_76" + ], + "output": "cat_12", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl" + }, + { + "node": "mul_104", + "inputs": [ + "cat_12", + "unsqueeze_16" + ], + "output": "mul_104", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "add_80", + "inputs": [ + "mul_103", + "mul_104" + ], + "output": "add_80", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl" + }, + { + "node": "cat_13", + "inputs": [ + "add_79", + "slice_71" + ], + "output": "cat_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl" + }, + { + "node": "cat_14", + "inputs": [ + "add_80", + "slice_73" + ], + "output": "cat_14", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl" + }, + { + "node": "cache_update_2", + "inputs": [ + "cat_14", + "position_ids" + ], + "output": "cache_update_2", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_14", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "cache_update_3", + "inputs": [ + "transpose_38", + "position_ids" + ], + "output": "cache_update_3", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_15", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "unsqueeze_17", + "inputs": [ + "cache_update_2" + ], + "output": "unsqueeze_17", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_4", + "inputs": [ + "unsqueeze_17" + ], + "output": "expand_4", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_11", + "inputs": [ + "expand_4" + ], + "output": "clone_11", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_4", + "inputs": [ + "clone_11" + ], + "output": "_unsafe_view_4", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "unsqueeze_18", + "inputs": [ + "cache_update_3" + ], + "output": "unsqueeze_18", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_5", + "inputs": [ + "unsqueeze_18" + ], + "output": "expand_5", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_12", + "inputs": [ + "expand_5" + ], + "output": "clone_12", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_5", + "inputs": [ + "clone_12" + ], + "output": "_unsafe_view_5", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "transpose_39", + "inputs": [ + "_unsafe_view_4" + ], + "output": "transpose_39", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl" + }, + { + "node": "matmul_3", + "inputs": [ + "cat_13", + "transpose_39" + ], + "output": "matmul_3", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl" + }, + { + "node": "mul_105", + "inputs": [ + "matmul_3" + ], + "output": "mul_105", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl" + }, + { + "node": "add_81", + "inputs": [ + "mul_105", + "attention_mask" + ], + "output": "add_81", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl" + }, + { + "node": "softmax_1", + "inputs": [ + "add_81" + ], + "output": "softmax_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f40727952ccdc71060abbdab6a141b0497a38b727d2f5b943fe826ce90497bd1.wgsl" + }, + { + "node": "_to_copy_108", + "inputs": [ + "softmax_1" + ], + "output": "_to_copy_108", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl" + }, + { + "node": "matmul_4", + "inputs": [ + "_to_copy_108", + "_unsafe_view_5" + ], + "output": "matmul_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl" + }, + { + "node": "transpose_40", + "inputs": [ + "matmul_4" + ], + "output": "transpose_40", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl" + }, + { + "node": "clone_13", + "inputs": [ + "transpose_40" + ], + "output": "clone_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_51", + "inputs": [ + "clone_13" + ], + "output": "view_51", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_7", + "inputs": [ + "_unsafe_view_3" + ], + "output": "sigmoid_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl" + }, + { + "node": "mul_106", + "inputs": [ + "view_51", + "sigmoid_7" + ], + "output": "mul_106", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/efd4310107ed3db32f1792f1360971ff7ce4234ab64fda03dcaaa9bff52bb912.wgsl" + }, + { + "node": "linear_58", + "inputs": [ + "mul_106", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight" + ], + "output": "linear_58", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_82", + "inputs": [ + "add_72", + "linear_58" + ], + "output": "add_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_109", + "inputs": [ + "add_82" + ], + "output": "_to_copy_109", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_26", + "inputs": [ + "_to_copy_109" + ], + "output": "pow_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_25", + "inputs": [ + "pow_26" + ], + "output": "mean_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_83", + "inputs": [ + "mean_25" + ], + "output": "add_83", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_37", + "inputs": [ + "add_83" + ], + "output": "rsqrt_37", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_107", + "inputs": [ + "_to_copy_109", + "rsqrt_37" + ], + "output": "mul_107", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_110", + "inputs": [ + "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + ], + "output": "_to_copy_110", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_84", + "inputs": [ + "_to_copy_110" + ], + "output": "add_84", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_108", + "inputs": [ + "mul_107", + "add_84" + ], + "output": "mul_108", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_19", + "inputs": [ + "mul_108" + ], + "output": "type_as_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_59", + "inputs": [ + "type_as_19", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight" + ], + "output": "linear_59", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_60", + "inputs": [ + "type_as_19", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_109", + "inputs": [ + "linear_59", + "linear_60" + ], + "output": "mul_109", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_61", + "inputs": [ + "mul_109", + "p_model_model_language_model_layers_7_mlp_down_proj_weight" + ], + "output": "linear_61", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_85", + "inputs": [ + "add_82", + "linear_61" + ], + "output": "add_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_111", + "inputs": [ + "add_85" + ], + "output": "_to_copy_111", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_27", + "inputs": [ + "_to_copy_111" + ], + "output": "pow_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_26", + "inputs": [ + "pow_27" + ], + "output": "mean_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_86", + "inputs": [ + "mean_26" + ], + "output": "add_86", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_38", + "inputs": [ + "add_86" + ], + "output": "rsqrt_38", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_110", + "inputs": [ + "_to_copy_111", + "rsqrt_38" + ], + "output": "mul_110", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_112", + "inputs": [ + "p_model_model_language_model_layers_8_input_layernorm_weight" + ], + "output": "_to_copy_112", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_87", + "inputs": [ + "_to_copy_112" + ], + "output": "add_87", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_111", + "inputs": [ + "mul_110", + "add_87" + ], + "output": "mul_111", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_20", + "inputs": [ + "mul_111" + ], + "output": "type_as_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_62", + "inputs": [ + "type_as_20", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_62", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_41", + "inputs": [ + "linear_62" + ], + "output": "transpose_41", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_63", + "inputs": [ + "type_as_20", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight" + ], + "output": "linear_63", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_52", + "inputs": [ + "linear_63" + ], + "output": "view_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_64", + "inputs": [ + "type_as_20", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight" + ], + "output": "linear_64", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_65", + "inputs": [ + "type_as_20", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight" + ], + "output": "linear_65", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_15", + "inputs": [ + "states_16", + "transpose_41" + ], + "output": "cat_15", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_78", + "inputs": [ + "cat_15" + ], + "output": "slice_78", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_16" + }, + { + "node": "squeeze_6", + "inputs": [ + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + ], + "output": "squeeze_6", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_19", + "inputs": [ + "squeeze_6" + ], + "output": "unsqueeze_19", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_6", + "inputs": [ + "cat_15", + "unsqueeze_19" + ], + "output": "conv1d_6", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_79", + "inputs": [ + "conv1d_6" + ], + "output": "slice_79", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_20", + "inputs": [ + "slice_79" + ], + "output": "silu_20", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_80", + "inputs": [ + "silu_20" + ], + "output": "slice_80", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_42", + "inputs": [ + "slice_80" + ], + "output": "transpose_42", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_81", + "inputs": [ + "transpose_42" + ], + "output": "slice_81", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_82", + "inputs": [ + "transpose_42" + ], + "output": "slice_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_83", + "inputs": [ + "transpose_42" + ], + "output": "slice_83", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_53", + "inputs": [ + "slice_81" + ], + "output": "view_53", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_54", + "inputs": [ + "slice_82" + ], + "output": "view_54", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_55", + "inputs": [ + "slice_83" + ], + "output": "view_55", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_8", + "inputs": [ + "linear_64" + ], + "output": "sigmoid_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_113", + "inputs": [ + "p_model_model_language_model_layers_8_linear_attn_a_log" + ], + "output": "_to_copy_113", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_6", + "inputs": [ + "_to_copy_113" + ], + "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": "_to_copy_114", + "inputs": [ + "linear_65" + ], + "output": "_to_copy_114", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_88", + "inputs": [ + "_to_copy_114", + "p_model_model_language_model_layers_8_linear_attn_dt_bias" + ], + "output": "add_88", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_53" + ], + "output": "mul_113", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_13", + "inputs": [ + "mul_113" + ], + "output": "sum_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_89", + "inputs": [ + "sum_13" + ], + "output": "add_89", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_39", + "inputs": [ + "add_89" + ], + "output": "rsqrt_39", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_114", + "inputs": [ + "view_53", + "rsqrt_39" + ], + "output": "mul_114", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_115", + "inputs": [ + "view_54" + ], + "output": "mul_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_14", + "inputs": [ + "mul_115" + ], + "output": "sum_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_90", + "inputs": [ + "sum_14" + ], + "output": "add_90", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_40", + "inputs": [ + "add_90" + ], + "output": "rsqrt_40", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_116", + "inputs": [ + "view_54", + "rsqrt_40" + ], + "output": "mul_116", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_115", + "inputs": [ + "mul_114" + ], + "output": "_to_copy_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_116", + "inputs": [ + "mul_116" + ], + "output": "_to_copy_116", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_117", + "inputs": [ + "view_55" + ], + "output": "_to_copy_117", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_118", + "inputs": [ + "sigmoid_8" + ], + "output": "_to_copy_118", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_6", + "inputs": [ + "_to_copy_115", + "_to_copy_116", + "_to_copy_117", + "mul_112", + "_to_copy_118", + "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": "clone_14", + "inputs": [ + "transpose_43" + ], + "output": "clone_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_119", + "inputs": [ + "clone_14" + ], + "output": "_to_copy_119", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_56", + "inputs": [ + "_to_copy_119" + ], + "output": "view_56", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_57", + "inputs": [ + "view_52" + ], + "output": "view_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_120", + "inputs": [ + "view_56" + ], + "output": "_to_copy_120", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_28", + "inputs": [ + "_to_copy_120" + ], + "output": "pow_28", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_27", + "inputs": [ + "pow_28" + ], + "output": "mean_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_91", + "inputs": [ + "mean_27" + ], + "output": "add_91", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_41", + "inputs": [ + "add_91" + ], + "output": "rsqrt_41", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_117", + "inputs": [ + "_to_copy_120", + "rsqrt_41" + ], + "output": "mul_117", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_121", + "inputs": [ + "mul_117" + ], + "output": "_to_copy_121", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_118", + "inputs": [ + "p_model_model_language_model_layers_8_linear_attn_norm_weight", + "_to_copy_121" + ], + "output": "mul_118", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_122", + "inputs": [ + "view_57" + ], + "output": "_to_copy_122", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_119", + "inputs": [ + "_to_copy_122", + "mul_118" + ], + "output": "mul_119", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_123", + "inputs": [ + "mul_119" + ], + "output": "_to_copy_123", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_58", + "inputs": [ + "_to_copy_123" + ], + "output": "view_58", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_66", + "inputs": [ + "view_58", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight" + ], + "output": "linear_66", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_92", + "inputs": [ + "add_85", + "linear_66" + ], + "output": "add_92", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_124", + "inputs": [ + "add_92" + ], + "output": "_to_copy_124", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_29", + "inputs": [ + "_to_copy_124" + ], + "output": "pow_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_28", + "inputs": [ + "pow_29" + ], + "output": "mean_28", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_93", + "inputs": [ + "mean_28" + ], + "output": "add_93", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_42", + "inputs": [ + "add_93" + ], + "output": "rsqrt_42", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_120", + "inputs": [ + "_to_copy_124", + "rsqrt_42" + ], + "output": "mul_120", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_125", + "inputs": [ + "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + ], + "output": "_to_copy_125", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_94", + "inputs": [ + "_to_copy_125" + ], + "output": "add_94", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_121", + "inputs": [ + "mul_120", + "add_94" + ], + "output": "mul_121", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_21", + "inputs": [ + "mul_121" + ], + "output": "type_as_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_67", + "inputs": [ + "type_as_21", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight" + ], + "output": "linear_67", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_68", + "inputs": [ + "type_as_21", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_122", + "inputs": [ + "linear_67", + "linear_68" + ], + "output": "mul_122", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_69", + "inputs": [ + "mul_122", + "p_model_model_language_model_layers_8_mlp_down_proj_weight" + ], + "output": "linear_69", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_95", + "inputs": [ + "add_92", + "linear_69" + ], + "output": "add_95", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_126", + "inputs": [ + "add_95" + ], + "output": "_to_copy_126", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_30", + "inputs": [ + "_to_copy_126" + ], + "output": "pow_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_29", + "inputs": [ + "pow_30" + ], + "output": "mean_29", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_96", + "inputs": [ + "mean_29" + ], + "output": "add_96", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_43", + "inputs": [ + "add_96" + ], + "output": "rsqrt_43", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_123", + "inputs": [ + "_to_copy_126", + "rsqrt_43" + ], + "output": "mul_123", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_127", + "inputs": [ + "p_model_model_language_model_layers_9_input_layernorm_weight" + ], + "output": "_to_copy_127", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_97", + "inputs": [ + "_to_copy_127" + ], + "output": "add_97", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_124", + "inputs": [ + "mul_123", + "add_97" + ], + "output": "mul_124", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_22", + "inputs": [ + "mul_124" + ], + "output": "type_as_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_70", + "inputs": [ + "type_as_22", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_70", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_44", + "inputs": [ + "linear_70" + ], + "output": "transpose_44", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_71", + "inputs": [ + "type_as_22", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight" + ], + "output": "linear_71", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_59", + "inputs": [ + "linear_71" + ], + "output": "view_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_72", + "inputs": [ + "type_as_22", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight" + ], + "output": "linear_72", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_73", + "inputs": [ + "type_as_22", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight" + ], + "output": "linear_73", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_16", + "inputs": [ + "states_18", + "transpose_44" + ], + "output": "cat_16", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_86", + "inputs": [ + "cat_16" + ], + "output": "slice_86", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_18" + }, + { + "node": "squeeze_7", + "inputs": [ + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + ], + "output": "squeeze_7", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_20", + "inputs": [ + "squeeze_7" + ], + "output": "unsqueeze_20", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_7", + "inputs": [ + "cat_16", + "unsqueeze_20" + ], + "output": "conv1d_7", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_87", + "inputs": [ + "conv1d_7" + ], + "output": "slice_87", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_23", + "inputs": [ + "slice_87" + ], + "output": "silu_23", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_88", + "inputs": [ + "silu_23" + ], + "output": "slice_88", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_45", + "inputs": [ + "slice_88" + ], + "output": "transpose_45", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_89", + "inputs": [ + "transpose_45" + ], + "output": "slice_89", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_90", + "inputs": [ + "transpose_45" + ], + "output": "slice_90", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_91", + "inputs": [ + "transpose_45" + ], + "output": "slice_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_60", + "inputs": [ + "slice_89" + ], + "output": "view_60", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_61", + "inputs": [ + "slice_90" + ], + "output": "view_61", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_62", + "inputs": [ + "slice_91" + ], + "output": "view_62", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_9", + "inputs": [ + "linear_72" + ], + "output": "sigmoid_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_128", + "inputs": [ + "p_model_model_language_model_layers_9_linear_attn_a_log" + ], + "output": "_to_copy_128", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_7", + "inputs": [ + "_to_copy_128" + ], + "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": "_to_copy_129", + "inputs": [ + "linear_73" + ], + "output": "_to_copy_129", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_98", + "inputs": [ + "_to_copy_129", + "p_model_model_language_model_layers_9_linear_attn_dt_bias" + ], + "output": "add_98", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_60" + ], + "output": "mul_126", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_15", + "inputs": [ + "mul_126" + ], + "output": "sum_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_99", + "inputs": [ + "sum_15" + ], + "output": "add_99", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_44", + "inputs": [ + "add_99" + ], + "output": "rsqrt_44", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_127", + "inputs": [ + "view_60", + "rsqrt_44" + ], + "output": "mul_127", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_128", + "inputs": [ + "view_61" + ], + "output": "mul_128", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_16", + "inputs": [ + "mul_128" + ], + "output": "sum_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_100", + "inputs": [ + "sum_16" + ], + "output": "add_100", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_45", + "inputs": [ + "add_100" + ], + "output": "rsqrt_45", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_129", + "inputs": [ + "view_61", + "rsqrt_45" + ], + "output": "mul_129", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_130", + "inputs": [ + "mul_127" + ], + "output": "_to_copy_130", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_131", + "inputs": [ + "mul_129" + ], + "output": "_to_copy_131", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_132", + "inputs": [ + "view_62" + ], + "output": "_to_copy_132", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_133", + "inputs": [ + "sigmoid_9" + ], + "output": "_to_copy_133", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_7", + "inputs": [ + "_to_copy_130", + "_to_copy_131", + "_to_copy_132", + "mul_125", + "_to_copy_133", + "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": "clone_15", + "inputs": [ + "transpose_46" + ], + "output": "clone_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_134", + "inputs": [ + "clone_15" + ], + "output": "_to_copy_134", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_63", + "inputs": [ + "_to_copy_134" + ], + "output": "view_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_64", + "inputs": [ + "view_59" + ], + "output": "view_64", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_135", + "inputs": [ + "view_63" + ], + "output": "_to_copy_135", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_31", + "inputs": [ + "_to_copy_135" + ], + "output": "pow_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_30", + "inputs": [ + "pow_31" + ], + "output": "mean_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_101", + "inputs": [ + "mean_30" + ], + "output": "add_101", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_46", + "inputs": [ + "add_101" + ], + "output": "rsqrt_46", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_130", + "inputs": [ + "_to_copy_135", + "rsqrt_46" + ], + "output": "mul_130", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_136", + "inputs": [ + "mul_130" + ], + "output": "_to_copy_136", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_131", + "inputs": [ + "p_model_model_language_model_layers_9_linear_attn_norm_weight", + "_to_copy_136" + ], + "output": "mul_131", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_137", + "inputs": [ + "view_64" + ], + "output": "_to_copy_137", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_132", + "inputs": [ + "_to_copy_137", + "mul_131" + ], + "output": "mul_132", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_138", + "inputs": [ + "mul_132" + ], + "output": "_to_copy_138", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_65", + "inputs": [ + "_to_copy_138" + ], + "output": "view_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_74", + "inputs": [ + "view_65", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + ], + "output": "linear_74", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_102", + "inputs": [ + "add_95", + "linear_74" + ], + "output": "add_102", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_139", + "inputs": [ + "add_102" + ], + "output": "_to_copy_139", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_32", + "inputs": [ + "_to_copy_139" + ], + "output": "pow_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_31", + "inputs": [ + "pow_32" + ], + "output": "mean_31", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_103", + "inputs": [ + "mean_31" + ], + "output": "add_103", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_47", + "inputs": [ + "add_103" + ], + "output": "rsqrt_47", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_133", + "inputs": [ + "_to_copy_139", + "rsqrt_47" + ], + "output": "mul_133", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_140", + "inputs": [ + "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + ], + "output": "_to_copy_140", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_104", + "inputs": [ + "_to_copy_140" + ], + "output": "add_104", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_134", + "inputs": [ + "mul_133", + "add_104" + ], + "output": "mul_134", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_23", + "inputs": [ + "mul_134" + ], + "output": "type_as_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_75", + "inputs": [ + "type_as_23", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight" + ], + "output": "linear_75", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_76", + "inputs": [ + "type_as_23", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_135", + "inputs": [ + "linear_75", + "linear_76" + ], + "output": "mul_135", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_77", + "inputs": [ + "mul_135", + "p_model_model_language_model_layers_9_mlp_down_proj_weight" + ], + "output": "linear_77", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_105", + "inputs": [ + "add_102", + "linear_77" + ], + "output": "add_105", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_141", + "inputs": [ + "add_105" + ], + "output": "_to_copy_141", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_33", + "inputs": [ + "_to_copy_141" + ], + "output": "pow_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_32", + "inputs": [ + "pow_33" + ], + "output": "mean_32", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_106", + "inputs": [ + "mean_32" + ], + "output": "add_106", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_48", + "inputs": [ + "add_106" + ], + "output": "rsqrt_48", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_136", + "inputs": [ + "_to_copy_141", + "rsqrt_48" + ], + "output": "mul_136", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_142", + "inputs": [ + "p_model_model_language_model_layers_10_input_layernorm_weight" + ], + "output": "_to_copy_142", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_107", + "inputs": [ + "_to_copy_142" + ], + "output": "add_107", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_137", + "inputs": [ + "mul_136", + "add_107" + ], + "output": "mul_137", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_24", + "inputs": [ + "mul_137" + ], + "output": "type_as_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_78", + "inputs": [ + "type_as_24", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_78", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_47", + "inputs": [ + "linear_78" + ], + "output": "transpose_47", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_79", + "inputs": [ + "type_as_24", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight" + ], + "output": "linear_79", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_66", + "inputs": [ + "linear_79" + ], + "output": "view_66", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_80", + "inputs": [ + "type_as_24", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight" + ], + "output": "linear_80", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_81", + "inputs": [ + "type_as_24", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight" + ], + "output": "linear_81", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_17", + "inputs": [ + "states_20", + "transpose_47" + ], + "output": "cat_17", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_94", + "inputs": [ + "cat_17" + ], + "output": "slice_94", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_20" + }, + { + "node": "squeeze_8", + "inputs": [ + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + ], + "output": "squeeze_8", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_21", + "inputs": [ + "squeeze_8" + ], + "output": "unsqueeze_21", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_8", + "inputs": [ + "cat_17", + "unsqueeze_21" + ], + "output": "conv1d_8", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_95", + "inputs": [ + "conv1d_8" + ], + "output": "slice_95", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_26", + "inputs": [ + "slice_95" + ], + "output": "silu_26", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_96", + "inputs": [ + "silu_26" + ], + "output": "slice_96", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_48", + "inputs": [ + "slice_96" + ], + "output": "transpose_48", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_97", + "inputs": [ + "transpose_48" + ], + "output": "slice_97", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_98", + "inputs": [ + "transpose_48" + ], + "output": "slice_98", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_99", + "inputs": [ + "transpose_48" + ], + "output": "slice_99", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_67", + "inputs": [ + "slice_97" + ], + "output": "view_67", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_68", + "inputs": [ + "slice_98" + ], + "output": "view_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_69", + "inputs": [ + "slice_99" + ], + "output": "view_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_10", + "inputs": [ + "linear_80" + ], + "output": "sigmoid_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_143", + "inputs": [ + "p_model_model_language_model_layers_10_linear_attn_a_log" + ], + "output": "_to_copy_143", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_8", + "inputs": [ + "_to_copy_143" + ], + "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": "_to_copy_144", + "inputs": [ + "linear_81" + ], + "output": "_to_copy_144", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_108", + "inputs": [ + "_to_copy_144", + "p_model_model_language_model_layers_10_linear_attn_dt_bias" + ], + "output": "add_108", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_67" + ], + "output": "mul_139", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_17", + "inputs": [ + "mul_139" + ], + "output": "sum_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_109", + "inputs": [ + "sum_17" + ], + "output": "add_109", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_49", + "inputs": [ + "add_109" + ], + "output": "rsqrt_49", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_140", + "inputs": [ + "view_67", + "rsqrt_49" + ], + "output": "mul_140", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_141", + "inputs": [ + "view_68" + ], + "output": "mul_141", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_18", + "inputs": [ + "mul_141" + ], + "output": "sum_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_110", + "inputs": [ + "sum_18" + ], + "output": "add_110", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_50", + "inputs": [ + "add_110" + ], + "output": "rsqrt_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_142", + "inputs": [ + "view_68", + "rsqrt_50" + ], + "output": "mul_142", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_145", + "inputs": [ + "mul_140" + ], + "output": "_to_copy_145", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_146", + "inputs": [ + "mul_142" + ], + "output": "_to_copy_146", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_147", + "inputs": [ + "view_69" + ], + "output": "_to_copy_147", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_148", + "inputs": [ + "sigmoid_10" + ], + "output": "_to_copy_148", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_8", + "inputs": [ + "_to_copy_145", + "_to_copy_146", + "_to_copy_147", + "mul_138", + "_to_copy_148", + "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": "clone_16", + "inputs": [ + "transpose_49" + ], + "output": "clone_16", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_149", + "inputs": [ + "clone_16" + ], + "output": "_to_copy_149", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_70", + "inputs": [ + "_to_copy_149" + ], + "output": "view_70", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_71", + "inputs": [ + "view_66" + ], + "output": "view_71", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_150", + "inputs": [ + "view_70" + ], + "output": "_to_copy_150", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_34", + "inputs": [ + "_to_copy_150" + ], + "output": "pow_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_33", + "inputs": [ + "pow_34" + ], + "output": "mean_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_111", + "inputs": [ + "mean_33" + ], + "output": "add_111", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_51", + "inputs": [ + "add_111" + ], + "output": "rsqrt_51", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_143", + "inputs": [ + "_to_copy_150", + "rsqrt_51" + ], + "output": "mul_143", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_151", + "inputs": [ + "mul_143" + ], + "output": "_to_copy_151", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_144", + "inputs": [ + "p_model_model_language_model_layers_10_linear_attn_norm_weight", + "_to_copy_151" + ], + "output": "mul_144", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_152", + "inputs": [ + "view_71" + ], + "output": "_to_copy_152", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_145", + "inputs": [ + "_to_copy_152", + "mul_144" + ], + "output": "mul_145", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_153", + "inputs": [ + "mul_145" + ], + "output": "_to_copy_153", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_72", + "inputs": [ + "_to_copy_153" + ], + "output": "view_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_82", + "inputs": [ + "view_72", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight" + ], + "output": "linear_82", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_112", + "inputs": [ + "add_105", + "linear_82" + ], + "output": "add_112", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_154", + "inputs": [ + "add_112" + ], + "output": "_to_copy_154", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_35", + "inputs": [ + "_to_copy_154" + ], + "output": "pow_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_34", + "inputs": [ + "pow_35" + ], + "output": "mean_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_113", + "inputs": [ + "mean_34" + ], + "output": "add_113", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_52", + "inputs": [ + "add_113" + ], + "output": "rsqrt_52", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_146", + "inputs": [ + "_to_copy_154", + "rsqrt_52" + ], + "output": "mul_146", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_155", + "inputs": [ + "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + ], + "output": "_to_copy_155", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_114", + "inputs": [ + "_to_copy_155" + ], + "output": "add_114", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_147", + "inputs": [ + "mul_146", + "add_114" + ], + "output": "mul_147", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_25", + "inputs": [ + "mul_147" + ], + "output": "type_as_25", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_83", + "inputs": [ + "type_as_25", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight" + ], + "output": "linear_83", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_84", + "inputs": [ + "type_as_25", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_148", + "inputs": [ + "linear_83", + "linear_84" + ], + "output": "mul_148", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_85", + "inputs": [ + "mul_148", + "p_model_model_language_model_layers_10_mlp_down_proj_weight" + ], + "output": "linear_85", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_115", + "inputs": [ + "add_112", + "linear_85" + ], + "output": "add_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_156", + "inputs": [ + "add_115" + ], + "output": "_to_copy_156", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_36", + "inputs": [ + "_to_copy_156" + ], + "output": "pow_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_35", + "inputs": [ + "pow_36" + ], + "output": "mean_35", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_116", + "inputs": [ + "mean_35" + ], + "output": "add_116", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_53", + "inputs": [ + "add_116" + ], + "output": "rsqrt_53", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_149", + "inputs": [ + "_to_copy_156", + "rsqrt_53" + ], + "output": "mul_149", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_157", + "inputs": [ + "p_model_model_language_model_layers_11_input_layernorm_weight" + ], + "output": "_to_copy_157", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_117", + "inputs": [ + "_to_copy_157" + ], + "output": "add_117", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_150", + "inputs": [ + "mul_149", + "add_117" + ], + "output": "mul_150", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_26", + "inputs": [ + "mul_150" + ], + "output": "type_as_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_86", + "inputs": [ + "type_as_26", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight" + ], + "output": "linear_86", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl" + }, + { + "node": "view_73", + "inputs": [ + "linear_86" + ], + "output": "view_73", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl" + }, + { + "node": "slice_102", + "inputs": [ + "view_73" + ], + "output": "slice_102", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c956fa9d3342793cbd859a6e47baea9cfafaac162fc2accf4b8f4728d0f604.wgsl" + }, + { + "node": "slice_103", + "inputs": [ + "view_73" + ], + "output": "slice_103", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl" + }, + { + "node": "clone_17", + "inputs": [ + "slice_103" + ], + "output": "clone_17", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_unsafe_view_6", + "inputs": [ + "clone_17" + ], + "output": "_unsafe_view_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_74", + "inputs": [ + "slice_102" + ], + "output": "view_74", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_158", + "inputs": [ + "view_74" + ], + "output": "_to_copy_158", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_37", + "inputs": [ + "_to_copy_158" + ], + "output": "pow_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_36", + "inputs": [ + "pow_37" + ], + "output": "mean_36", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl" + }, + { + "node": "add_118", + "inputs": [ + "mean_36" + ], + "output": "add_118", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl" + }, + { + "node": "rsqrt_54", + "inputs": [ + "add_118" + ], + "output": "rsqrt_54", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl" + }, + { + "node": "mul_151", + "inputs": [ + "_to_copy_158", + "rsqrt_54" + ], + "output": "mul_151", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl" + }, + { + "node": "_to_copy_159", + "inputs": [ + "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + ], + "output": "_to_copy_159", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_119", + "inputs": [ + "_to_copy_159" + ], + "output": "add_119", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_152", + "inputs": [ + "mul_151", + "add_119" + ], + "output": "mul_152", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl" + }, + { + "node": "type_as_27", + "inputs": [ + "mul_152" + ], + "output": "type_as_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "transpose_50", + "inputs": [ + "type_as_27" + ], + "output": "transpose_50", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa1c36207dbf73b89a207971c82a8ed3c252e4b5057c44d46f9646ecce99da76.wgsl" + }, + { + "node": "linear_87", + "inputs": [ + "type_as_26", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight" + ], + "output": "linear_87", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_75", + "inputs": [ + "linear_87" + ], + "output": "view_75", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "_to_copy_160", + "inputs": [ + "view_75" + ], + "output": "_to_copy_160", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "pow_38", + "inputs": [ + "_to_copy_160" + ], + "output": "pow_38", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl" + }, + { + "node": "mean_37", + "inputs": [ + "pow_38" + ], + "output": "mean_37", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl" + }, + { + "node": "add_120", + "inputs": [ + "mean_37" + ], + "output": "add_120", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl" + }, + { + "node": "rsqrt_55", + "inputs": [ + "add_120" + ], + "output": "rsqrt_55", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl" + }, + { + "node": "mul_153", + "inputs": [ + "_to_copy_160", + "rsqrt_55" + ], + "output": "mul_153", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl" + }, + { + "node": "_to_copy_161", + "inputs": [ + "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + ], + "output": "_to_copy_161", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_121", + "inputs": [ + "_to_copy_161" + ], + "output": "add_121", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_154", + "inputs": [ + "mul_153", + "add_121" + ], + "output": "mul_154", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl" + }, + { + "node": "type_as_28", + "inputs": [ + "mul_154" + ], + "output": "type_as_28", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f8a4395bd0356791917d7ef0469f9b7d983d33d4b7ad044d66852ead8a86971b.wgsl" + }, + { + "node": "transpose_51", + "inputs": [ + "type_as_28" + ], + "output": "transpose_51", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "linear_88", + "inputs": [ + "type_as_26", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight" + ], + "output": "linear_88", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_76", + "inputs": [ + "linear_88" + ], + "output": "view_76", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "transpose_52", + "inputs": [ + "view_76" + ], + "output": "transpose_52", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "unsqueeze_22", + "inputs": [ + "_to_copy_1" + ], + "output": "unsqueeze_22", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "unsqueeze_23", + "inputs": [ + "_to_copy_2" + ], + "output": "unsqueeze_23", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "slice_104", + "inputs": [ + "transpose_50" + ], + "output": "slice_104", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl" + }, + { + "node": "slice_105", + "inputs": [ + "transpose_50" + ], + "output": "slice_105", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl" + }, + { + "node": "slice_106", + "inputs": [ + "transpose_51" + ], + "output": "slice_106", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e697b1bc29cd0d08bba0c6a455a7e4b02a3c2e2535304faff2c04b6976617c7c.wgsl" + }, + { + "node": "slice_107", + "inputs": [ + "transpose_51" + ], + "output": "slice_107", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl" + }, + { + "node": "mul_155", + "inputs": [ + "slice_104", + "unsqueeze_22" + ], + "output": "mul_155", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "slice_108", + "inputs": [ + "slice_104" + ], + "output": "slice_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl" + }, + { + "node": "slice_109", + "inputs": [ + "slice_104" + ], + "output": "slice_109", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl" + }, + { + "node": "neg_13", + "inputs": [ + "slice_109" + ], + "output": "neg_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl" + }, + { + "node": "cat_18", + "inputs": [ + "neg_13", + "slice_108" + ], + "output": "cat_18", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl" + }, + { + "node": "mul_156", + "inputs": [ + "cat_18", + "unsqueeze_23" + ], + "output": "mul_156", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "add_122", + "inputs": [ + "mul_155", + "mul_156" + ], + "output": "add_122", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl" + }, + { + "node": "mul_157", + "inputs": [ + "slice_106", + "unsqueeze_22" + ], + "output": "mul_157", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "slice_110", + "inputs": [ + "slice_106" + ], + "output": "slice_110", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/fd2084f233c932d1160371a9a522dbc8abdf416285eef54e02e18e4738d12675.wgsl" + }, + { + "node": "slice_111", + "inputs": [ + "slice_106" + ], + "output": "slice_111", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl" + }, + { + "node": "neg_14", + "inputs": [ + "slice_111" + ], + "output": "neg_14", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl" + }, + { + "node": "cat_19", + "inputs": [ + "neg_14", + "slice_110" + ], + "output": "cat_19", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl" + }, + { + "node": "mul_158", + "inputs": [ + "cat_19", + "unsqueeze_23" + ], + "output": "mul_158", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "add_123", + "inputs": [ + "mul_157", + "mul_158" + ], + "output": "add_123", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl" + }, + { + "node": "cat_20", + "inputs": [ + "add_122", + "slice_105" + ], + "output": "cat_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl" + }, + { + "node": "cat_21", + "inputs": [ + "add_123", + "slice_107" + ], + "output": "cat_21", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl" + }, + { + "node": "cache_update_4", + "inputs": [ + "cat_21", + "position_ids" + ], + "output": "cache_update_4", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_22", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "cache_update_5", + "inputs": [ + "transpose_52", + "position_ids" + ], + "output": "cache_update_5", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_23", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "unsqueeze_24", + "inputs": [ + "cache_update_4" + ], + "output": "unsqueeze_24", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_6", + "inputs": [ + "unsqueeze_24" + ], + "output": "expand_6", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_18", + "inputs": [ + "expand_6" + ], + "output": "clone_18", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_7", + "inputs": [ + "clone_18" + ], + "output": "_unsafe_view_7", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "unsqueeze_25", + "inputs": [ + "cache_update_5" + ], + "output": "unsqueeze_25", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_7", + "inputs": [ + "unsqueeze_25" + ], + "output": "expand_7", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_19", + "inputs": [ + "expand_7" + ], + "output": "clone_19", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_8", + "inputs": [ + "clone_19" + ], + "output": "_unsafe_view_8", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "transpose_53", + "inputs": [ + "_unsafe_view_7" + ], + "output": "transpose_53", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl" + }, + { + "node": "matmul_5", + "inputs": [ + "cat_20", + "transpose_53" + ], + "output": "matmul_5", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl" + }, + { + "node": "mul_159", + "inputs": [ + "matmul_5" + ], + "output": "mul_159", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl" + }, + { + "node": "add_124", + "inputs": [ + "mul_159", + "attention_mask" + ], + "output": "add_124", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl" + }, + { + "node": "softmax_2", + "inputs": [ + "add_124" + ], + "output": "softmax_2", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f40727952ccdc71060abbdab6a141b0497a38b727d2f5b943fe826ce90497bd1.wgsl" + }, + { + "node": "_to_copy_162", + "inputs": [ + "softmax_2" + ], + "output": "_to_copy_162", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl" + }, + { + "node": "matmul_6", + "inputs": [ + "_to_copy_162", + "_unsafe_view_8" + ], + "output": "matmul_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl" + }, + { + "node": "transpose_54", + "inputs": [ + "matmul_6" + ], + "output": "transpose_54", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl" + }, + { + "node": "clone_20", + "inputs": [ + "transpose_54" + ], + "output": "clone_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_77", + "inputs": [ + "clone_20" + ], + "output": "view_77", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_11", + "inputs": [ + "_unsafe_view_6" + ], + "output": "sigmoid_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl" + }, + { + "node": "mul_160", + "inputs": [ + "view_77", + "sigmoid_11" + ], + "output": "mul_160", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/efd4310107ed3db32f1792f1360971ff7ce4234ab64fda03dcaaa9bff52bb912.wgsl" + }, + { + "node": "linear_89", + "inputs": [ + "mul_160", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight" + ], + "output": "linear_89", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_125", + "inputs": [ + "add_115", + "linear_89" + ], + "output": "add_125", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_163", + "inputs": [ + "add_125" + ], + "output": "_to_copy_163", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_39", + "inputs": [ + "_to_copy_163" + ], + "output": "pow_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_38", + "inputs": [ + "pow_39" + ], + "output": "mean_38", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_126", + "inputs": [ + "mean_38" + ], + "output": "add_126", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_56", + "inputs": [ + "add_126" + ], + "output": "rsqrt_56", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_161", + "inputs": [ + "_to_copy_163", + "rsqrt_56" + ], + "output": "mul_161", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_164", + "inputs": [ + "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + ], + "output": "_to_copy_164", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_127", + "inputs": [ + "_to_copy_164" + ], + "output": "add_127", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_162", + "inputs": [ + "mul_161", + "add_127" + ], + "output": "mul_162", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_29", + "inputs": [ + "mul_162" + ], + "output": "type_as_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_90", + "inputs": [ + "type_as_29", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight" + ], + "output": "linear_90", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_91", + "inputs": [ + "type_as_29", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_163", + "inputs": [ + "linear_90", + "linear_91" + ], + "output": "mul_163", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_92", + "inputs": [ + "mul_163", + "p_model_model_language_model_layers_11_mlp_down_proj_weight" + ], + "output": "linear_92", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_128", + "inputs": [ + "add_125", + "linear_92" + ], + "output": "add_128", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_165", + "inputs": [ + "add_128" + ], + "output": "_to_copy_165", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_40", + "inputs": [ + "_to_copy_165" + ], + "output": "pow_40", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_39", + "inputs": [ + "pow_40" + ], + "output": "mean_39", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_129", + "inputs": [ + "mean_39" + ], + "output": "add_129", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_57", + "inputs": [ + "add_129" + ], + "output": "rsqrt_57", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_164", + "inputs": [ + "_to_copy_165", + "rsqrt_57" + ], + "output": "mul_164", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_166", + "inputs": [ + "p_model_model_language_model_layers_12_input_layernorm_weight" + ], + "output": "_to_copy_166", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_130", + "inputs": [ + "_to_copy_166" + ], + "output": "add_130", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_165", + "inputs": [ + "mul_164", + "add_130" + ], + "output": "mul_165", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_30", + "inputs": [ + "mul_165" + ], + "output": "type_as_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_93", + "inputs": [ + "type_as_30", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_93", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_55", + "inputs": [ + "linear_93" + ], + "output": "transpose_55", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_94", + "inputs": [ + "type_as_30", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight" + ], + "output": "linear_94", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_78", + "inputs": [ + "linear_94" + ], + "output": "view_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_95", + "inputs": [ + "type_as_30", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight" + ], + "output": "linear_95", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_96", + "inputs": [ + "type_as_30", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight" + ], + "output": "linear_96", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_22", + "inputs": [ + "states_24", + "transpose_55" + ], + "output": "cat_22", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_112", + "inputs": [ + "cat_22" + ], + "output": "slice_112", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_24" + }, + { + "node": "squeeze_9", + "inputs": [ + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + ], + "output": "squeeze_9", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_26", + "inputs": [ + "squeeze_9" + ], + "output": "unsqueeze_26", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_9", + "inputs": [ + "cat_22", + "unsqueeze_26" + ], + "output": "conv1d_9", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_113", + "inputs": [ + "conv1d_9" + ], + "output": "slice_113", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_30", + "inputs": [ + "slice_113" + ], + "output": "silu_30", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_114", + "inputs": [ + "silu_30" + ], + "output": "slice_114", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_56", + "inputs": [ + "slice_114" + ], + "output": "transpose_56", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_115", + "inputs": [ + "transpose_56" + ], + "output": "slice_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_116", + "inputs": [ + "transpose_56" + ], + "output": "slice_116", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_117", + "inputs": [ + "transpose_56" + ], + "output": "slice_117", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_79", + "inputs": [ + "slice_115" + ], + "output": "view_79", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_80", + "inputs": [ + "slice_116" + ], + "output": "view_80", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_81", + "inputs": [ + "slice_117" + ], + "output": "view_81", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_12", + "inputs": [ + "linear_95" + ], + "output": "sigmoid_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_167", + "inputs": [ + "p_model_model_language_model_layers_12_linear_attn_a_log" + ], + "output": "_to_copy_167", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_9", + "inputs": [ + "_to_copy_167" + ], + "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": "_to_copy_168", + "inputs": [ + "linear_96" + ], + "output": "_to_copy_168", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_131", + "inputs": [ + "_to_copy_168", + "p_model_model_language_model_layers_12_linear_attn_dt_bias" + ], + "output": "add_131", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_79" + ], + "output": "mul_167", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_19", + "inputs": [ + "mul_167" + ], + "output": "sum_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_132", + "inputs": [ + "sum_19" + ], + "output": "add_132", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_58", + "inputs": [ + "add_132" + ], + "output": "rsqrt_58", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_168", + "inputs": [ + "view_79", + "rsqrt_58" + ], + "output": "mul_168", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_169", + "inputs": [ + "view_80" + ], + "output": "mul_169", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_20", + "inputs": [ + "mul_169" + ], + "output": "sum_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_133", + "inputs": [ + "sum_20" + ], + "output": "add_133", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_59", + "inputs": [ + "add_133" + ], + "output": "rsqrt_59", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_170", + "inputs": [ + "view_80", + "rsqrt_59" + ], + "output": "mul_170", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_169", + "inputs": [ + "mul_168" + ], + "output": "_to_copy_169", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_170", + "inputs": [ + "mul_170" + ], + "output": "_to_copy_170", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_171", + "inputs": [ + "view_81" + ], + "output": "_to_copy_171", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_172", + "inputs": [ + "sigmoid_12" + ], + "output": "_to_copy_172", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_9", + "inputs": [ + "_to_copy_169", + "_to_copy_170", + "_to_copy_171", + "mul_166", + "_to_copy_172", + "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": "clone_21", + "inputs": [ + "transpose_57" + ], + "output": "clone_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_173", + "inputs": [ + "clone_21" + ], + "output": "_to_copy_173", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_82", + "inputs": [ + "_to_copy_173" + ], + "output": "view_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_83", + "inputs": [ + "view_78" + ], + "output": "view_83", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_174", + "inputs": [ + "view_82" + ], + "output": "_to_copy_174", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_41", + "inputs": [ + "_to_copy_174" + ], + "output": "pow_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_40", + "inputs": [ + "pow_41" + ], + "output": "mean_40", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_134", + "inputs": [ + "mean_40" + ], + "output": "add_134", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_60", + "inputs": [ + "add_134" + ], + "output": "rsqrt_60", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_171", + "inputs": [ + "_to_copy_174", + "rsqrt_60" + ], + "output": "mul_171", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_175", + "inputs": [ + "mul_171" + ], + "output": "_to_copy_175", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_172", + "inputs": [ + "p_model_model_language_model_layers_12_linear_attn_norm_weight", + "_to_copy_175" + ], + "output": "mul_172", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_176", + "inputs": [ + "view_83" + ], + "output": "_to_copy_176", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_173", + "inputs": [ + "_to_copy_176", + "mul_172" + ], + "output": "mul_173", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_177", + "inputs": [ + "mul_173" + ], + "output": "_to_copy_177", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_84", + "inputs": [ + "_to_copy_177" + ], + "output": "view_84", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_97", + "inputs": [ + "view_84", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight" + ], + "output": "linear_97", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_135", + "inputs": [ + "add_128", + "linear_97" + ], + "output": "add_135", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_178", + "inputs": [ + "add_135" + ], + "output": "_to_copy_178", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_42", + "inputs": [ + "_to_copy_178" + ], + "output": "pow_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_41", + "inputs": [ + "pow_42" + ], + "output": "mean_41", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_136", + "inputs": [ + "mean_41" + ], + "output": "add_136", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_61", + "inputs": [ + "add_136" + ], + "output": "rsqrt_61", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_174", + "inputs": [ + "_to_copy_178", + "rsqrt_61" + ], + "output": "mul_174", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_179", + "inputs": [ + "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + ], + "output": "_to_copy_179", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_137", + "inputs": [ + "_to_copy_179" + ], + "output": "add_137", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_175", + "inputs": [ + "mul_174", + "add_137" + ], + "output": "mul_175", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_31", + "inputs": [ + "mul_175" + ], + "output": "type_as_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_98", + "inputs": [ + "type_as_31", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight" + ], + "output": "linear_98", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_99", + "inputs": [ + "type_as_31", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_176", + "inputs": [ + "linear_98", + "linear_99" + ], + "output": "mul_176", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_100", + "inputs": [ + "mul_176", + "p_model_model_language_model_layers_12_mlp_down_proj_weight" + ], + "output": "linear_100", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_138", + "inputs": [ + "add_135", + "linear_100" + ], + "output": "add_138", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_180", + "inputs": [ + "add_138" + ], + "output": "_to_copy_180", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_43", + "inputs": [ + "_to_copy_180" + ], + "output": "pow_43", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_42", + "inputs": [ + "pow_43" + ], + "output": "mean_42", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_139", + "inputs": [ + "mean_42" + ], + "output": "add_139", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_62", + "inputs": [ + "add_139" + ], + "output": "rsqrt_62", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_177", + "inputs": [ + "_to_copy_180", + "rsqrt_62" + ], + "output": "mul_177", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_181", + "inputs": [ + "p_model_model_language_model_layers_13_input_layernorm_weight" + ], + "output": "_to_copy_181", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_140", + "inputs": [ + "_to_copy_181" + ], + "output": "add_140", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_178", + "inputs": [ + "mul_177", + "add_140" + ], + "output": "mul_178", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_32", + "inputs": [ + "mul_178" + ], + "output": "type_as_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_101", + "inputs": [ + "type_as_32", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_101", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_58", + "inputs": [ + "linear_101" + ], + "output": "transpose_58", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_102", + "inputs": [ + "type_as_32", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight" + ], + "output": "linear_102", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_85", + "inputs": [ + "linear_102" + ], + "output": "view_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_103", + "inputs": [ + "type_as_32", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight" + ], + "output": "linear_103", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_104", + "inputs": [ + "type_as_32", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight" + ], + "output": "linear_104", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_23", + "inputs": [ + "states_26", + "transpose_58" + ], + "output": "cat_23", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_120", + "inputs": [ + "cat_23" + ], + "output": "slice_120", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_26" + }, + { + "node": "squeeze_10", + "inputs": [ + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + ], + "output": "squeeze_10", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_27", + "inputs": [ + "squeeze_10" + ], + "output": "unsqueeze_27", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_10", + "inputs": [ + "cat_23", + "unsqueeze_27" + ], + "output": "conv1d_10", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_121", + "inputs": [ + "conv1d_10" + ], + "output": "slice_121", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_33", + "inputs": [ + "slice_121" + ], + "output": "silu_33", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_122", + "inputs": [ + "silu_33" + ], + "output": "slice_122", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_59", + "inputs": [ + "slice_122" + ], + "output": "transpose_59", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_123", + "inputs": [ + "transpose_59" + ], + "output": "slice_123", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_124", + "inputs": [ + "transpose_59" + ], + "output": "slice_124", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_125", + "inputs": [ + "transpose_59" + ], + "output": "slice_125", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_86", + "inputs": [ + "slice_123" + ], + "output": "view_86", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_87", + "inputs": [ + "slice_124" + ], + "output": "view_87", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_88", + "inputs": [ + "slice_125" + ], + "output": "view_88", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_13", + "inputs": [ + "linear_103" + ], + "output": "sigmoid_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_182", + "inputs": [ + "p_model_model_language_model_layers_13_linear_attn_a_log" + ], + "output": "_to_copy_182", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_10", + "inputs": [ + "_to_copy_182" + ], + "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": "_to_copy_183", + "inputs": [ + "linear_104" + ], + "output": "_to_copy_183", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_141", + "inputs": [ + "_to_copy_183", + "p_model_model_language_model_layers_13_linear_attn_dt_bias" + ], + "output": "add_141", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_86" + ], + "output": "mul_180", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_21", + "inputs": [ + "mul_180" + ], + "output": "sum_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_142", + "inputs": [ + "sum_21" + ], + "output": "add_142", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_63", + "inputs": [ + "add_142" + ], + "output": "rsqrt_63", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_181", + "inputs": [ + "view_86", + "rsqrt_63" + ], + "output": "mul_181", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_182", + "inputs": [ + "view_87" + ], + "output": "mul_182", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_22", + "inputs": [ + "mul_182" + ], + "output": "sum_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_143", + "inputs": [ + "sum_22" + ], + "output": "add_143", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_64", + "inputs": [ + "add_143" + ], + "output": "rsqrt_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_183", + "inputs": [ + "view_87", + "rsqrt_64" + ], + "output": "mul_183", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_184", + "inputs": [ + "mul_181" + ], + "output": "_to_copy_184", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_185", + "inputs": [ + "mul_183" + ], + "output": "_to_copy_185", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_186", + "inputs": [ + "view_88" + ], + "output": "_to_copy_186", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_187", + "inputs": [ + "sigmoid_13" + ], + "output": "_to_copy_187", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_10", + "inputs": [ + "_to_copy_184", + "_to_copy_185", + "_to_copy_186", + "mul_179", + "_to_copy_187", + "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": "clone_22", + "inputs": [ + "transpose_60" + ], + "output": "clone_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_188", + "inputs": [ + "clone_22" + ], + "output": "_to_copy_188", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_89", + "inputs": [ + "_to_copy_188" + ], + "output": "view_89", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_90", + "inputs": [ + "view_85" + ], + "output": "view_90", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_189", + "inputs": [ + "view_89" + ], + "output": "_to_copy_189", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_44", + "inputs": [ + "_to_copy_189" + ], + "output": "pow_44", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_43", + "inputs": [ + "pow_44" + ], + "output": "mean_43", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_144", + "inputs": [ + "mean_43" + ], + "output": "add_144", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_65", + "inputs": [ + "add_144" + ], + "output": "rsqrt_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_184", + "inputs": [ + "_to_copy_189", + "rsqrt_65" + ], + "output": "mul_184", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_190", + "inputs": [ + "mul_184" + ], + "output": "_to_copy_190", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_185", + "inputs": [ + "p_model_model_language_model_layers_13_linear_attn_norm_weight", + "_to_copy_190" + ], + "output": "mul_185", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_191", + "inputs": [ + "view_90" + ], + "output": "_to_copy_191", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_186", + "inputs": [ + "_to_copy_191", + "mul_185" + ], + "output": "mul_186", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_192", + "inputs": [ + "mul_186" + ], + "output": "_to_copy_192", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_91", + "inputs": [ + "_to_copy_192" + ], + "output": "view_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_105", + "inputs": [ + "view_91", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight" + ], + "output": "linear_105", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_145", + "inputs": [ + "add_138", + "linear_105" + ], + "output": "add_145", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_193", + "inputs": [ + "add_145" + ], + "output": "_to_copy_193", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_45", + "inputs": [ + "_to_copy_193" + ], + "output": "pow_45", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_44", + "inputs": [ + "pow_45" + ], + "output": "mean_44", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_146", + "inputs": [ + "mean_44" + ], + "output": "add_146", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_66", + "inputs": [ + "add_146" + ], + "output": "rsqrt_66", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_187", + "inputs": [ + "_to_copy_193", + "rsqrt_66" + ], + "output": "mul_187", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_194", + "inputs": [ + "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + ], + "output": "_to_copy_194", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_147", + "inputs": [ + "_to_copy_194" + ], + "output": "add_147", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_188", + "inputs": [ + "mul_187", + "add_147" + ], + "output": "mul_188", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_33", + "inputs": [ + "mul_188" + ], + "output": "type_as_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_106", + "inputs": [ + "type_as_33", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight" + ], + "output": "linear_106", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_107", + "inputs": [ + "type_as_33", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_189", + "inputs": [ + "linear_106", + "linear_107" + ], + "output": "mul_189", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_108", + "inputs": [ + "mul_189", + "p_model_model_language_model_layers_13_mlp_down_proj_weight" + ], + "output": "linear_108", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_148", + "inputs": [ + "add_145", + "linear_108" + ], + "output": "add_148", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_195", + "inputs": [ + "add_148" + ], + "output": "_to_copy_195", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_46", + "inputs": [ + "_to_copy_195" + ], + "output": "pow_46", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_45", + "inputs": [ + "pow_46" + ], + "output": "mean_45", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_149", + "inputs": [ + "mean_45" + ], + "output": "add_149", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_67", + "inputs": [ + "add_149" + ], + "output": "rsqrt_67", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_190", + "inputs": [ + "_to_copy_195", + "rsqrt_67" + ], + "output": "mul_190", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_196", + "inputs": [ + "p_model_model_language_model_layers_14_input_layernorm_weight" + ], + "output": "_to_copy_196", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_150", + "inputs": [ + "_to_copy_196" + ], + "output": "add_150", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_191", + "inputs": [ + "mul_190", + "add_150" + ], + "output": "mul_191", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_34", + "inputs": [ + "mul_191" + ], + "output": "type_as_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_109", + "inputs": [ + "type_as_34", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_109", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_61", + "inputs": [ + "linear_109" + ], + "output": "transpose_61", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_110", + "inputs": [ + "type_as_34", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight" + ], + "output": "linear_110", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_92", + "inputs": [ + "linear_110" + ], + "output": "view_92", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_111", + "inputs": [ + "type_as_34", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight" + ], + "output": "linear_111", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_112", + "inputs": [ + "type_as_34", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight" + ], + "output": "linear_112", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_24", + "inputs": [ + "states_28", + "transpose_61" + ], + "output": "cat_24", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_128", + "inputs": [ + "cat_24" + ], + "output": "slice_128", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_28" + }, + { + "node": "squeeze_11", + "inputs": [ + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + ], + "output": "squeeze_11", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_28", + "inputs": [ + "squeeze_11" + ], + "output": "unsqueeze_28", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_11", + "inputs": [ + "cat_24", + "unsqueeze_28" + ], + "output": "conv1d_11", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_129", + "inputs": [ + "conv1d_11" + ], + "output": "slice_129", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_36", + "inputs": [ + "slice_129" + ], + "output": "silu_36", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_130", + "inputs": [ + "silu_36" + ], + "output": "slice_130", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_62", + "inputs": [ + "slice_130" + ], + "output": "transpose_62", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_131", + "inputs": [ + "transpose_62" + ], + "output": "slice_131", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_132", + "inputs": [ + "transpose_62" + ], + "output": "slice_132", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_133", + "inputs": [ + "transpose_62" + ], + "output": "slice_133", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_93", + "inputs": [ + "slice_131" + ], + "output": "view_93", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_94", + "inputs": [ + "slice_132" + ], + "output": "view_94", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_95", + "inputs": [ + "slice_133" + ], + "output": "view_95", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_14", + "inputs": [ + "linear_111" + ], + "output": "sigmoid_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_197", + "inputs": [ + "p_model_model_language_model_layers_14_linear_attn_a_log" + ], + "output": "_to_copy_197", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_11", + "inputs": [ + "_to_copy_197" + ], + "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": "_to_copy_198", + "inputs": [ + "linear_112" + ], + "output": "_to_copy_198", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_151", + "inputs": [ + "_to_copy_198", + "p_model_model_language_model_layers_14_linear_attn_dt_bias" + ], + "output": "add_151", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_93" + ], + "output": "mul_193", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_23", + "inputs": [ + "mul_193" + ], + "output": "sum_23", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_152", + "inputs": [ + "sum_23" + ], + "output": "add_152", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_68", + "inputs": [ + "add_152" + ], + "output": "rsqrt_68", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_194", + "inputs": [ + "view_93", + "rsqrt_68" + ], + "output": "mul_194", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_195", + "inputs": [ + "view_94" + ], + "output": "mul_195", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_24", + "inputs": [ + "mul_195" + ], + "output": "sum_24", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_153", + "inputs": [ + "sum_24" + ], + "output": "add_153", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_69", + "inputs": [ + "add_153" + ], + "output": "rsqrt_69", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_196", + "inputs": [ + "view_94", + "rsqrt_69" + ], + "output": "mul_196", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_199", + "inputs": [ + "mul_194" + ], + "output": "_to_copy_199", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_200", + "inputs": [ + "mul_196" + ], + "output": "_to_copy_200", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_201", + "inputs": [ + "view_95" + ], + "output": "_to_copy_201", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_202", + "inputs": [ + "sigmoid_14" + ], + "output": "_to_copy_202", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_11", + "inputs": [ + "_to_copy_199", + "_to_copy_200", + "_to_copy_201", + "mul_192", + "_to_copy_202", + "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": "clone_23", + "inputs": [ + "transpose_63" + ], + "output": "clone_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_203", + "inputs": [ + "clone_23" + ], + "output": "_to_copy_203", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_96", + "inputs": [ + "_to_copy_203" + ], + "output": "view_96", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_97", + "inputs": [ + "view_92" + ], + "output": "view_97", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_204", + "inputs": [ + "view_96" + ], + "output": "_to_copy_204", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_47", + "inputs": [ + "_to_copy_204" + ], + "output": "pow_47", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_46", + "inputs": [ + "pow_47" + ], + "output": "mean_46", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_154", + "inputs": [ + "mean_46" + ], + "output": "add_154", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_70", + "inputs": [ + "add_154" + ], + "output": "rsqrt_70", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_197", + "inputs": [ + "_to_copy_204", + "rsqrt_70" + ], + "output": "mul_197", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_205", + "inputs": [ + "mul_197" + ], + "output": "_to_copy_205", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_198", + "inputs": [ + "p_model_model_language_model_layers_14_linear_attn_norm_weight", + "_to_copy_205" + ], + "output": "mul_198", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_206", + "inputs": [ + "view_97" + ], + "output": "_to_copy_206", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_199", + "inputs": [ + "_to_copy_206", + "mul_198" + ], + "output": "mul_199", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_207", + "inputs": [ + "mul_199" + ], + "output": "_to_copy_207", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_98", + "inputs": [ + "_to_copy_207" + ], + "output": "view_98", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_113", + "inputs": [ + "view_98", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight" + ], + "output": "linear_113", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_155", + "inputs": [ + "add_148", + "linear_113" + ], + "output": "add_155", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_208", + "inputs": [ + "add_155" + ], + "output": "_to_copy_208", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_48", + "inputs": [ + "_to_copy_208" + ], + "output": "pow_48", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_47", + "inputs": [ + "pow_48" + ], + "output": "mean_47", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_156", + "inputs": [ + "mean_47" + ], + "output": "add_156", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_71", + "inputs": [ + "add_156" + ], + "output": "rsqrt_71", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_200", + "inputs": [ + "_to_copy_208", + "rsqrt_71" + ], + "output": "mul_200", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_209", + "inputs": [ + "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + ], + "output": "_to_copy_209", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_157", + "inputs": [ + "_to_copy_209" + ], + "output": "add_157", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_201", + "inputs": [ + "mul_200", + "add_157" + ], + "output": "mul_201", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_35", + "inputs": [ + "mul_201" + ], + "output": "type_as_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_114", + "inputs": [ + "type_as_35", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight" + ], + "output": "linear_114", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_115", + "inputs": [ + "type_as_35", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_202", + "inputs": [ + "linear_114", + "linear_115" + ], + "output": "mul_202", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_116", + "inputs": [ + "mul_202", + "p_model_model_language_model_layers_14_mlp_down_proj_weight" + ], + "output": "linear_116", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_158", + "inputs": [ + "add_155", + "linear_116" + ], + "output": "add_158", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_210", + "inputs": [ + "add_158" + ], + "output": "_to_copy_210", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_49", + "inputs": [ + "_to_copy_210" + ], + "output": "pow_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_48", + "inputs": [ + "pow_49" + ], + "output": "mean_48", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_159", + "inputs": [ + "mean_48" + ], + "output": "add_159", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_72", + "inputs": [ + "add_159" + ], + "output": "rsqrt_72", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_203", + "inputs": [ + "_to_copy_210", + "rsqrt_72" + ], + "output": "mul_203", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_211", + "inputs": [ + "p_model_model_language_model_layers_15_input_layernorm_weight" + ], + "output": "_to_copy_211", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_160", + "inputs": [ + "_to_copy_211" + ], + "output": "add_160", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_204", + "inputs": [ + "mul_203", + "add_160" + ], + "output": "mul_204", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_36", + "inputs": [ + "mul_204" + ], + "output": "type_as_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_117", + "inputs": [ + "type_as_36", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight" + ], + "output": "linear_117", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl" + }, + { + "node": "view_99", + "inputs": [ + "linear_117" + ], + "output": "view_99", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl" + }, + { + "node": "slice_136", + "inputs": [ + "view_99" + ], + "output": "slice_136", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c956fa9d3342793cbd859a6e47baea9cfafaac162fc2accf4b8f4728d0f604.wgsl" + }, + { + "node": "slice_137", + "inputs": [ + "view_99" + ], + "output": "slice_137", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl" + }, + { + "node": "clone_24", + "inputs": [ + "slice_137" + ], + "output": "clone_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_unsafe_view_9", + "inputs": [ + "clone_24" + ], + "output": "_unsafe_view_9", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_100", + "inputs": [ + "slice_136" + ], + "output": "view_100", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_212", + "inputs": [ + "view_100" + ], + "output": "_to_copy_212", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_50", + "inputs": [ + "_to_copy_212" + ], + "output": "pow_50", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_49", + "inputs": [ + "pow_50" + ], + "output": "mean_49", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl" + }, + { + "node": "add_161", + "inputs": [ + "mean_49" + ], + "output": "add_161", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl" + }, + { + "node": "rsqrt_73", + "inputs": [ + "add_161" + ], + "output": "rsqrt_73", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl" + }, + { + "node": "mul_205", + "inputs": [ + "_to_copy_212", + "rsqrt_73" + ], + "output": "mul_205", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl" + }, + { + "node": "_to_copy_213", + "inputs": [ + "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + ], + "output": "_to_copy_213", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_162", + "inputs": [ + "_to_copy_213" + ], + "output": "add_162", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_206", + "inputs": [ + "mul_205", + "add_162" + ], + "output": "mul_206", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl" + }, + { + "node": "type_as_37", + "inputs": [ + "mul_206" + ], + "output": "type_as_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "transpose_64", + "inputs": [ + "type_as_37" + ], + "output": "transpose_64", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa1c36207dbf73b89a207971c82a8ed3c252e4b5057c44d46f9646ecce99da76.wgsl" + }, + { + "node": "linear_118", + "inputs": [ + "type_as_36", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight" + ], + "output": "linear_118", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_101", + "inputs": [ + "linear_118" + ], + "output": "view_101", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "_to_copy_214", + "inputs": [ + "view_101" + ], + "output": "_to_copy_214", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "pow_51", + "inputs": [ + "_to_copy_214" + ], + "output": "pow_51", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl" + }, + { + "node": "mean_50", + "inputs": [ + "pow_51" + ], + "output": "mean_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl" + }, + { + "node": "add_163", + "inputs": [ + "mean_50" + ], + "output": "add_163", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl" + }, + { + "node": "rsqrt_74", + "inputs": [ + "add_163" + ], + "output": "rsqrt_74", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl" + }, + { + "node": "mul_207", + "inputs": [ + "_to_copy_214", + "rsqrt_74" + ], + "output": "mul_207", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl" + }, + { + "node": "_to_copy_215", + "inputs": [ + "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + ], + "output": "_to_copy_215", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_164", + "inputs": [ + "_to_copy_215" + ], + "output": "add_164", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_208", + "inputs": [ + "mul_207", + "add_164" + ], + "output": "mul_208", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl" + }, + { + "node": "type_as_38", + "inputs": [ + "mul_208" + ], + "output": "type_as_38", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f8a4395bd0356791917d7ef0469f9b7d983d33d4b7ad044d66852ead8a86971b.wgsl" + }, + { + "node": "transpose_65", + "inputs": [ + "type_as_38" + ], + "output": "transpose_65", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "linear_119", + "inputs": [ + "type_as_36", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight" + ], + "output": "linear_119", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_102", + "inputs": [ + "linear_119" + ], + "output": "view_102", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "transpose_66", + "inputs": [ + "view_102" + ], + "output": "transpose_66", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "unsqueeze_29", + "inputs": [ + "_to_copy_1" + ], + "output": "unsqueeze_29", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "unsqueeze_30", + "inputs": [ + "_to_copy_2" + ], + "output": "unsqueeze_30", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "slice_138", + "inputs": [ + "transpose_64" + ], + "output": "slice_138", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl" + }, + { + "node": "slice_139", + "inputs": [ + "transpose_64" + ], + "output": "slice_139", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl" + }, + { + "node": "slice_140", + "inputs": [ + "transpose_65" + ], + "output": "slice_140", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e697b1bc29cd0d08bba0c6a455a7e4b02a3c2e2535304faff2c04b6976617c7c.wgsl" + }, + { + "node": "slice_141", + "inputs": [ + "transpose_65" + ], + "output": "slice_141", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl" + }, + { + "node": "mul_209", + "inputs": [ + "slice_138", + "unsqueeze_29" + ], + "output": "mul_209", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "slice_142", + "inputs": [ + "slice_138" + ], + "output": "slice_142", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl" + }, + { + "node": "slice_143", + "inputs": [ + "slice_138" + ], + "output": "slice_143", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl" + }, + { + "node": "neg_18", + "inputs": [ + "slice_143" + ], + "output": "neg_18", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl" + }, + { + "node": "cat_25", + "inputs": [ + "neg_18", + "slice_142" + ], + "output": "cat_25", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl" + }, + { + "node": "mul_210", + "inputs": [ + "cat_25", + "unsqueeze_30" + ], + "output": "mul_210", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "add_165", + "inputs": [ + "mul_209", + "mul_210" + ], + "output": "add_165", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl" + }, + { + "node": "mul_211", + "inputs": [ + "slice_140", + "unsqueeze_29" + ], + "output": "mul_211", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "slice_144", + "inputs": [ + "slice_140" + ], + "output": "slice_144", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/fd2084f233c932d1160371a9a522dbc8abdf416285eef54e02e18e4738d12675.wgsl" + }, + { + "node": "slice_145", + "inputs": [ + "slice_140" + ], + "output": "slice_145", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl" + }, + { + "node": "neg_19", + "inputs": [ + "slice_145" + ], + "output": "neg_19", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl" + }, + { + "node": "cat_26", + "inputs": [ + "neg_19", + "slice_144" + ], + "output": "cat_26", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl" + }, + { + "node": "mul_212", + "inputs": [ + "cat_26", + "unsqueeze_30" + ], + "output": "mul_212", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "add_166", + "inputs": [ + "mul_211", + "mul_212" + ], + "output": "add_166", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl" + }, + { + "node": "cat_27", + "inputs": [ + "add_165", + "slice_139" + ], + "output": "cat_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl" + }, + { + "node": "cat_28", + "inputs": [ + "add_166", + "slice_141" + ], + "output": "cat_28", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl" + }, + { + "node": "cache_update_6", + "inputs": [ + "cat_28", + "position_ids" + ], + "output": "cache_update_6", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_30", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "cache_update_7", + "inputs": [ + "transpose_66", + "position_ids" + ], + "output": "cache_update_7", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_31", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "unsqueeze_31", + "inputs": [ + "cache_update_6" + ], + "output": "unsqueeze_31", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_8", + "inputs": [ + "unsqueeze_31" + ], + "output": "expand_8", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_25", + "inputs": [ + "expand_8" + ], + "output": "clone_25", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_10", + "inputs": [ + "clone_25" + ], + "output": "_unsafe_view_10", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "unsqueeze_32", + "inputs": [ + "cache_update_7" + ], + "output": "unsqueeze_32", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_9", + "inputs": [ + "unsqueeze_32" + ], + "output": "expand_9", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_26", + "inputs": [ + "expand_9" + ], + "output": "clone_26", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_11", + "inputs": [ + "clone_26" + ], + "output": "_unsafe_view_11", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "transpose_67", + "inputs": [ + "_unsafe_view_10" + ], + "output": "transpose_67", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl" + }, + { + "node": "matmul_7", + "inputs": [ + "cat_27", + "transpose_67" + ], + "output": "matmul_7", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl" + }, + { + "node": "mul_213", + "inputs": [ + "matmul_7" + ], + "output": "mul_213", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl" + }, + { + "node": "add_167", + "inputs": [ + "mul_213", + "attention_mask" + ], + "output": "add_167", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl" + }, + { + "node": "softmax_3", + "inputs": [ + "add_167" + ], + "output": "softmax_3", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f40727952ccdc71060abbdab6a141b0497a38b727d2f5b943fe826ce90497bd1.wgsl" + }, + { + "node": "_to_copy_216", + "inputs": [ + "softmax_3" + ], + "output": "_to_copy_216", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl" + }, + { + "node": "matmul_8", + "inputs": [ + "_to_copy_216", + "_unsafe_view_11" + ], + "output": "matmul_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl" + }, + { + "node": "transpose_68", + "inputs": [ + "matmul_8" + ], + "output": "transpose_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl" + }, + { + "node": "clone_27", + "inputs": [ + "transpose_68" + ], + "output": "clone_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_103", + "inputs": [ + "clone_27" + ], + "output": "view_103", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_15", + "inputs": [ + "_unsafe_view_9" + ], + "output": "sigmoid_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl" + }, + { + "node": "mul_214", + "inputs": [ + "view_103", + "sigmoid_15" + ], + "output": "mul_214", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/efd4310107ed3db32f1792f1360971ff7ce4234ab64fda03dcaaa9bff52bb912.wgsl" + }, + { + "node": "linear_120", + "inputs": [ + "mul_214", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight" + ], + "output": "linear_120", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_168", + "inputs": [ + "add_158", + "linear_120" + ], + "output": "add_168", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_217", + "inputs": [ + "add_168" + ], + "output": "_to_copy_217", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_52", + "inputs": [ + "_to_copy_217" + ], + "output": "pow_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_51", + "inputs": [ + "pow_52" + ], + "output": "mean_51", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_169", + "inputs": [ + "mean_51" + ], + "output": "add_169", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_75", + "inputs": [ + "add_169" + ], + "output": "rsqrt_75", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_215", + "inputs": [ + "_to_copy_217", + "rsqrt_75" + ], + "output": "mul_215", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_218", + "inputs": [ + "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + ], + "output": "_to_copy_218", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_170", + "inputs": [ + "_to_copy_218" + ], + "output": "add_170", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_216", + "inputs": [ + "mul_215", + "add_170" + ], + "output": "mul_216", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_39", + "inputs": [ + "mul_216" + ], + "output": "type_as_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_121", + "inputs": [ + "type_as_39", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight" + ], + "output": "linear_121", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_122", + "inputs": [ + "type_as_39", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_217", + "inputs": [ + "linear_121", + "linear_122" + ], + "output": "mul_217", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_123", + "inputs": [ + "mul_217", + "p_model_model_language_model_layers_15_mlp_down_proj_weight" + ], + "output": "linear_123", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_171", + "inputs": [ + "add_168", + "linear_123" + ], + "output": "add_171", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_219", + "inputs": [ + "add_171" + ], + "output": "_to_copy_219", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_53", + "inputs": [ + "_to_copy_219" + ], + "output": "pow_53", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_52", + "inputs": [ + "pow_53" + ], + "output": "mean_52", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_172", + "inputs": [ + "mean_52" + ], + "output": "add_172", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_76", + "inputs": [ + "add_172" + ], + "output": "rsqrt_76", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_218", + "inputs": [ + "_to_copy_219", + "rsqrt_76" + ], + "output": "mul_218", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_220", + "inputs": [ + "p_model_model_language_model_layers_16_input_layernorm_weight" + ], + "output": "_to_copy_220", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_173", + "inputs": [ + "_to_copy_220" + ], + "output": "add_173", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_219", + "inputs": [ + "mul_218", + "add_173" + ], + "output": "mul_219", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_40", + "inputs": [ + "mul_219" + ], + "output": "type_as_40", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_124", + "inputs": [ + "type_as_40", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_124", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_69", + "inputs": [ + "linear_124" + ], + "output": "transpose_69", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_125", + "inputs": [ + "type_as_40", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight" + ], + "output": "linear_125", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_104", + "inputs": [ + "linear_125" + ], + "output": "view_104", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_126", + "inputs": [ + "type_as_40", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight" + ], + "output": "linear_126", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_127", + "inputs": [ + "type_as_40", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight" + ], + "output": "linear_127", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_29", + "inputs": [ + "states_32", + "transpose_69" + ], + "output": "cat_29", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_146", + "inputs": [ + "cat_29" + ], + "output": "slice_146", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_32" + }, + { + "node": "squeeze_12", + "inputs": [ + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + ], + "output": "squeeze_12", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_33", + "inputs": [ + "squeeze_12" + ], + "output": "unsqueeze_33", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_12", + "inputs": [ + "cat_29", + "unsqueeze_33" + ], + "output": "conv1d_12", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_147", + "inputs": [ + "conv1d_12" + ], + "output": "slice_147", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_40", + "inputs": [ + "slice_147" + ], + "output": "silu_40", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_148", + "inputs": [ + "silu_40" + ], + "output": "slice_148", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_70", + "inputs": [ + "slice_148" + ], + "output": "transpose_70", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_149", + "inputs": [ + "transpose_70" + ], + "output": "slice_149", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_150", + "inputs": [ + "transpose_70" + ], + "output": "slice_150", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_151", + "inputs": [ + "transpose_70" + ], + "output": "slice_151", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_105", + "inputs": [ + "slice_149" + ], + "output": "view_105", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_106", + "inputs": [ + "slice_150" + ], + "output": "view_106", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_107", + "inputs": [ + "slice_151" + ], + "output": "view_107", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_16", + "inputs": [ + "linear_126" + ], + "output": "sigmoid_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_221", + "inputs": [ + "p_model_model_language_model_layers_16_linear_attn_a_log" + ], + "output": "_to_copy_221", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_12", + "inputs": [ + "_to_copy_221" + ], + "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": "_to_copy_222", + "inputs": [ + "linear_127" + ], + "output": "_to_copy_222", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_174", + "inputs": [ + "_to_copy_222", + "p_model_model_language_model_layers_16_linear_attn_dt_bias" + ], + "output": "add_174", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_105" + ], + "output": "mul_221", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_25", + "inputs": [ + "mul_221" + ], + "output": "sum_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_175", + "inputs": [ + "sum_25" + ], + "output": "add_175", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_77", + "inputs": [ + "add_175" + ], + "output": "rsqrt_77", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_222", + "inputs": [ + "view_105", + "rsqrt_77" + ], + "output": "mul_222", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_223", + "inputs": [ + "view_106" + ], + "output": "mul_223", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_26", + "inputs": [ + "mul_223" + ], + "output": "sum_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_176", + "inputs": [ + "sum_26" + ], + "output": "add_176", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_78", + "inputs": [ + "add_176" + ], + "output": "rsqrt_78", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_224", + "inputs": [ + "view_106", + "rsqrt_78" + ], + "output": "mul_224", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_223", + "inputs": [ + "mul_222" + ], + "output": "_to_copy_223", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_224", + "inputs": [ + "mul_224" + ], + "output": "_to_copy_224", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_225", + "inputs": [ + "view_107" + ], + "output": "_to_copy_225", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_226", + "inputs": [ + "sigmoid_16" + ], + "output": "_to_copy_226", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_12", + "inputs": [ + "_to_copy_223", + "_to_copy_224", + "_to_copy_225", + "mul_220", + "_to_copy_226", + "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": "clone_28", + "inputs": [ + "transpose_71" + ], + "output": "clone_28", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_227", + "inputs": [ + "clone_28" + ], + "output": "_to_copy_227", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_108", + "inputs": [ + "_to_copy_227" + ], + "output": "view_108", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_109", + "inputs": [ + "view_104" + ], + "output": "view_109", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_228", + "inputs": [ + "view_108" + ], + "output": "_to_copy_228", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_54", + "inputs": [ + "_to_copy_228" + ], + "output": "pow_54", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_53", + "inputs": [ + "pow_54" + ], + "output": "mean_53", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_177", + "inputs": [ + "mean_53" + ], + "output": "add_177", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_79", + "inputs": [ + "add_177" + ], + "output": "rsqrt_79", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_225", + "inputs": [ + "_to_copy_228", + "rsqrt_79" + ], + "output": "mul_225", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_229", + "inputs": [ + "mul_225" + ], + "output": "_to_copy_229", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_226", + "inputs": [ + "p_model_model_language_model_layers_16_linear_attn_norm_weight", + "_to_copy_229" + ], + "output": "mul_226", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_230", + "inputs": [ + "view_109" + ], + "output": "_to_copy_230", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_227", + "inputs": [ + "_to_copy_230", + "mul_226" + ], + "output": "mul_227", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_231", + "inputs": [ + "mul_227" + ], + "output": "_to_copy_231", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_110", + "inputs": [ + "_to_copy_231" + ], + "output": "view_110", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_128", + "inputs": [ + "view_110", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight" + ], + "output": "linear_128", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_178", + "inputs": [ + "add_171", + "linear_128" + ], + "output": "add_178", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_232", + "inputs": [ + "add_178" + ], + "output": "_to_copy_232", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_55", + "inputs": [ + "_to_copy_232" + ], + "output": "pow_55", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_54", + "inputs": [ + "pow_55" + ], + "output": "mean_54", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_179", + "inputs": [ + "mean_54" + ], + "output": "add_179", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_80", + "inputs": [ + "add_179" + ], + "output": "rsqrt_80", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_228", + "inputs": [ + "_to_copy_232", + "rsqrt_80" + ], + "output": "mul_228", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_233", + "inputs": [ + "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + ], + "output": "_to_copy_233", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_180", + "inputs": [ + "_to_copy_233" + ], + "output": "add_180", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_229", + "inputs": [ + "mul_228", + "add_180" + ], + "output": "mul_229", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_41", + "inputs": [ + "mul_229" + ], + "output": "type_as_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_129", + "inputs": [ + "type_as_41", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight" + ], + "output": "linear_129", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_130", + "inputs": [ + "type_as_41", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_230", + "inputs": [ + "linear_129", + "linear_130" + ], + "output": "mul_230", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_131", + "inputs": [ + "mul_230", + "p_model_model_language_model_layers_16_mlp_down_proj_weight" + ], + "output": "linear_131", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_181", + "inputs": [ + "add_178", + "linear_131" + ], + "output": "add_181", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_234", + "inputs": [ + "add_181" + ], + "output": "_to_copy_234", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_56", + "inputs": [ + "_to_copy_234" + ], + "output": "pow_56", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_55", + "inputs": [ + "pow_56" + ], + "output": "mean_55", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_182", + "inputs": [ + "mean_55" + ], + "output": "add_182", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_81", + "inputs": [ + "add_182" + ], + "output": "rsqrt_81", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_231", + "inputs": [ + "_to_copy_234", + "rsqrt_81" + ], + "output": "mul_231", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_235", + "inputs": [ + "p_model_model_language_model_layers_17_input_layernorm_weight" + ], + "output": "_to_copy_235", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_183", + "inputs": [ + "_to_copy_235" + ], + "output": "add_183", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_232", + "inputs": [ + "mul_231", + "add_183" + ], + "output": "mul_232", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_42", + "inputs": [ + "mul_232" + ], + "output": "type_as_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_132", + "inputs": [ + "type_as_42", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_132", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_72", + "inputs": [ + "linear_132" + ], + "output": "transpose_72", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_133", + "inputs": [ + "type_as_42", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight" + ], + "output": "linear_133", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_111", + "inputs": [ + "linear_133" + ], + "output": "view_111", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_134", + "inputs": [ + "type_as_42", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight" + ], + "output": "linear_134", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_135", + "inputs": [ + "type_as_42", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight" + ], + "output": "linear_135", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_30", + "inputs": [ + "states_34", + "transpose_72" + ], + "output": "cat_30", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_154", + "inputs": [ + "cat_30" + ], + "output": "slice_154", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_34" + }, + { + "node": "squeeze_13", + "inputs": [ + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + ], + "output": "squeeze_13", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_34", + "inputs": [ + "squeeze_13" + ], + "output": "unsqueeze_34", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_13", + "inputs": [ + "cat_30", + "unsqueeze_34" + ], + "output": "conv1d_13", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_155", + "inputs": [ + "conv1d_13" + ], + "output": "slice_155", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_43", + "inputs": [ + "slice_155" + ], + "output": "silu_43", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_156", + "inputs": [ + "silu_43" + ], + "output": "slice_156", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_73", + "inputs": [ + "slice_156" + ], + "output": "transpose_73", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_157", + "inputs": [ + "transpose_73" + ], + "output": "slice_157", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_158", + "inputs": [ + "transpose_73" + ], + "output": "slice_158", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_159", + "inputs": [ + "transpose_73" + ], + "output": "slice_159", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_112", + "inputs": [ + "slice_157" + ], + "output": "view_112", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_113", + "inputs": [ + "slice_158" + ], + "output": "view_113", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_114", + "inputs": [ + "slice_159" + ], + "output": "view_114", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_17", + "inputs": [ + "linear_134" + ], + "output": "sigmoid_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_236", + "inputs": [ + "p_model_model_language_model_layers_17_linear_attn_a_log" + ], + "output": "_to_copy_236", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_13", + "inputs": [ + "_to_copy_236" + ], + "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": "_to_copy_237", + "inputs": [ + "linear_135" + ], + "output": "_to_copy_237", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_184", + "inputs": [ + "_to_copy_237", + "p_model_model_language_model_layers_17_linear_attn_dt_bias" + ], + "output": "add_184", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_112" + ], + "output": "mul_234", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_27", + "inputs": [ + "mul_234" + ], + "output": "sum_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_185", + "inputs": [ + "sum_27" + ], + "output": "add_185", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_82", + "inputs": [ + "add_185" + ], + "output": "rsqrt_82", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_235", + "inputs": [ + "view_112", + "rsqrt_82" + ], + "output": "mul_235", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_236", + "inputs": [ + "view_113" + ], + "output": "mul_236", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_28", + "inputs": [ + "mul_236" + ], + "output": "sum_28", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_186", + "inputs": [ + "sum_28" + ], + "output": "add_186", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_83", + "inputs": [ + "add_186" + ], + "output": "rsqrt_83", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_237", + "inputs": [ + "view_113", + "rsqrt_83" + ], + "output": "mul_237", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_238", + "inputs": [ + "mul_235" + ], + "output": "_to_copy_238", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_239", + "inputs": [ + "mul_237" + ], + "output": "_to_copy_239", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_240", + "inputs": [ + "view_114" + ], + "output": "_to_copy_240", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_241", + "inputs": [ + "sigmoid_17" + ], + "output": "_to_copy_241", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_13", + "inputs": [ + "_to_copy_238", + "_to_copy_239", + "_to_copy_240", + "mul_233", + "_to_copy_241", + "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": "clone_29", + "inputs": [ + "transpose_74" + ], + "output": "clone_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_242", + "inputs": [ + "clone_29" + ], + "output": "_to_copy_242", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_115", + "inputs": [ + "_to_copy_242" + ], + "output": "view_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_116", + "inputs": [ + "view_111" + ], + "output": "view_116", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_243", + "inputs": [ + "view_115" + ], + "output": "_to_copy_243", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_57", + "inputs": [ + "_to_copy_243" + ], + "output": "pow_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_56", + "inputs": [ + "pow_57" + ], + "output": "mean_56", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_187", + "inputs": [ + "mean_56" + ], + "output": "add_187", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_84", + "inputs": [ + "add_187" + ], + "output": "rsqrt_84", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_238", + "inputs": [ + "_to_copy_243", + "rsqrt_84" + ], + "output": "mul_238", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_244", + "inputs": [ + "mul_238" + ], + "output": "_to_copy_244", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_239", + "inputs": [ + "p_model_model_language_model_layers_17_linear_attn_norm_weight", + "_to_copy_244" + ], + "output": "mul_239", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_245", + "inputs": [ + "view_116" + ], + "output": "_to_copy_245", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_240", + "inputs": [ + "_to_copy_245", + "mul_239" + ], + "output": "mul_240", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_246", + "inputs": [ + "mul_240" + ], + "output": "_to_copy_246", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_117", + "inputs": [ + "_to_copy_246" + ], + "output": "view_117", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_136", + "inputs": [ + "view_117", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight" + ], + "output": "linear_136", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_188", + "inputs": [ + "add_181", + "linear_136" + ], + "output": "add_188", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_247", + "inputs": [ + "add_188" + ], + "output": "_to_copy_247", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_58", + "inputs": [ + "_to_copy_247" + ], + "output": "pow_58", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_57", + "inputs": [ + "pow_58" + ], + "output": "mean_57", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_189", + "inputs": [ + "mean_57" + ], + "output": "add_189", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_85", + "inputs": [ + "add_189" + ], + "output": "rsqrt_85", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_241", + "inputs": [ + "_to_copy_247", + "rsqrt_85" + ], + "output": "mul_241", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_248", + "inputs": [ + "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + ], + "output": "_to_copy_248", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_190", + "inputs": [ + "_to_copy_248" + ], + "output": "add_190", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_242", + "inputs": [ + "mul_241", + "add_190" + ], + "output": "mul_242", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_43", + "inputs": [ + "mul_242" + ], + "output": "type_as_43", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_137", + "inputs": [ + "type_as_43", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight" + ], + "output": "linear_137", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_138", + "inputs": [ + "type_as_43", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_243", + "inputs": [ + "linear_137", + "linear_138" + ], + "output": "mul_243", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_139", + "inputs": [ + "mul_243", + "p_model_model_language_model_layers_17_mlp_down_proj_weight" + ], + "output": "linear_139", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_191", + "inputs": [ + "add_188", + "linear_139" + ], + "output": "add_191", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_249", + "inputs": [ + "add_191" + ], + "output": "_to_copy_249", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_59", + "inputs": [ + "_to_copy_249" + ], + "output": "pow_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_58", + "inputs": [ + "pow_59" + ], + "output": "mean_58", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_192", + "inputs": [ + "mean_58" + ], + "output": "add_192", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_86", + "inputs": [ + "add_192" + ], + "output": "rsqrt_86", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_244", + "inputs": [ + "_to_copy_249", + "rsqrt_86" + ], + "output": "mul_244", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_250", + "inputs": [ + "p_model_model_language_model_layers_18_input_layernorm_weight" + ], + "output": "_to_copy_250", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_193", + "inputs": [ + "_to_copy_250" + ], + "output": "add_193", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_245", + "inputs": [ + "mul_244", + "add_193" + ], + "output": "mul_245", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_44", + "inputs": [ + "mul_245" + ], + "output": "type_as_44", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_140", + "inputs": [ + "type_as_44", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_140", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_75", + "inputs": [ + "linear_140" + ], + "output": "transpose_75", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_141", + "inputs": [ + "type_as_44", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight" + ], + "output": "linear_141", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_118", + "inputs": [ + "linear_141" + ], + "output": "view_118", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_142", + "inputs": [ + "type_as_44", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight" + ], + "output": "linear_142", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_143", + "inputs": [ + "type_as_44", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight" + ], + "output": "linear_143", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_31", + "inputs": [ + "states_36", + "transpose_75" + ], + "output": "cat_31", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_162", + "inputs": [ + "cat_31" + ], + "output": "slice_162", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_36" + }, + { + "node": "squeeze_14", + "inputs": [ + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + ], + "output": "squeeze_14", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_35", + "inputs": [ + "squeeze_14" + ], + "output": "unsqueeze_35", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_14", + "inputs": [ + "cat_31", + "unsqueeze_35" + ], + "output": "conv1d_14", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_163", + "inputs": [ + "conv1d_14" + ], + "output": "slice_163", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_46", + "inputs": [ + "slice_163" + ], + "output": "silu_46", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_164", + "inputs": [ + "silu_46" + ], + "output": "slice_164", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_76", + "inputs": [ + "slice_164" + ], + "output": "transpose_76", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_165", + "inputs": [ + "transpose_76" + ], + "output": "slice_165", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_166", + "inputs": [ + "transpose_76" + ], + "output": "slice_166", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_167", + "inputs": [ + "transpose_76" + ], + "output": "slice_167", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_119", + "inputs": [ + "slice_165" + ], + "output": "view_119", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_120", + "inputs": [ + "slice_166" + ], + "output": "view_120", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_121", + "inputs": [ + "slice_167" + ], + "output": "view_121", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_18", + "inputs": [ + "linear_142" + ], + "output": "sigmoid_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_251", + "inputs": [ + "p_model_model_language_model_layers_18_linear_attn_a_log" + ], + "output": "_to_copy_251", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_14", + "inputs": [ + "_to_copy_251" + ], + "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": "_to_copy_252", + "inputs": [ + "linear_143" + ], + "output": "_to_copy_252", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_194", + "inputs": [ + "_to_copy_252", + "p_model_model_language_model_layers_18_linear_attn_dt_bias" + ], + "output": "add_194", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_119" + ], + "output": "mul_247", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_29", + "inputs": [ + "mul_247" + ], + "output": "sum_29", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_195", + "inputs": [ + "sum_29" + ], + "output": "add_195", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_87", + "inputs": [ + "add_195" + ], + "output": "rsqrt_87", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_248", + "inputs": [ + "view_119", + "rsqrt_87" + ], + "output": "mul_248", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_249", + "inputs": [ + "view_120" + ], + "output": "mul_249", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_30", + "inputs": [ + "mul_249" + ], + "output": "sum_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_196", + "inputs": [ + "sum_30" + ], + "output": "add_196", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_88", + "inputs": [ + "add_196" + ], + "output": "rsqrt_88", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_250", + "inputs": [ + "view_120", + "rsqrt_88" + ], + "output": "mul_250", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_253", + "inputs": [ + "mul_248" + ], + "output": "_to_copy_253", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_254", + "inputs": [ + "mul_250" + ], + "output": "_to_copy_254", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_255", + "inputs": [ + "view_121" + ], + "output": "_to_copy_255", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_256", + "inputs": [ + "sigmoid_18" + ], + "output": "_to_copy_256", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_14", + "inputs": [ + "_to_copy_253", + "_to_copy_254", + "_to_copy_255", + "mul_246", + "_to_copy_256", + "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": "clone_30", + "inputs": [ + "transpose_77" + ], + "output": "clone_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_257", + "inputs": [ + "clone_30" + ], + "output": "_to_copy_257", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_122", + "inputs": [ + "_to_copy_257" + ], + "output": "view_122", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_123", + "inputs": [ + "view_118" + ], + "output": "view_123", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_258", + "inputs": [ + "view_122" + ], + "output": "_to_copy_258", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_60", + "inputs": [ + "_to_copy_258" + ], + "output": "pow_60", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_59", + "inputs": [ + "pow_60" + ], + "output": "mean_59", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_197", + "inputs": [ + "mean_59" + ], + "output": "add_197", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_89", + "inputs": [ + "add_197" + ], + "output": "rsqrt_89", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_251", + "inputs": [ + "_to_copy_258", + "rsqrt_89" + ], + "output": "mul_251", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_259", + "inputs": [ + "mul_251" + ], + "output": "_to_copy_259", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_252", + "inputs": [ + "p_model_model_language_model_layers_18_linear_attn_norm_weight", + "_to_copy_259" + ], + "output": "mul_252", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_260", + "inputs": [ + "view_123" + ], + "output": "_to_copy_260", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_253", + "inputs": [ + "_to_copy_260", + "mul_252" + ], + "output": "mul_253", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_261", + "inputs": [ + "mul_253" + ], + "output": "_to_copy_261", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_124", + "inputs": [ + "_to_copy_261" + ], + "output": "view_124", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_144", + "inputs": [ + "view_124", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight" + ], + "output": "linear_144", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_198", + "inputs": [ + "add_191", + "linear_144" + ], + "output": "add_198", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_262", + "inputs": [ + "add_198" + ], + "output": "_to_copy_262", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_61", + "inputs": [ + "_to_copy_262" + ], + "output": "pow_61", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_60", + "inputs": [ + "pow_61" + ], + "output": "mean_60", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_199", + "inputs": [ + "mean_60" + ], + "output": "add_199", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_90", + "inputs": [ + "add_199" + ], + "output": "rsqrt_90", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_254", + "inputs": [ + "_to_copy_262", + "rsqrt_90" + ], + "output": "mul_254", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_263", + "inputs": [ + "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + ], + "output": "_to_copy_263", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_200", + "inputs": [ + "_to_copy_263" + ], + "output": "add_200", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_255", + "inputs": [ + "mul_254", + "add_200" + ], + "output": "mul_255", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_45", + "inputs": [ + "mul_255" + ], + "output": "type_as_45", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_145", + "inputs": [ + "type_as_45", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight" + ], + "output": "linear_145", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_146", + "inputs": [ + "type_as_45", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_256", + "inputs": [ + "linear_145", + "linear_146" + ], + "output": "mul_256", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_147", + "inputs": [ + "mul_256", + "p_model_model_language_model_layers_18_mlp_down_proj_weight" + ], + "output": "linear_147", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_201", + "inputs": [ + "add_198", + "linear_147" + ], + "output": "add_201", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_264", + "inputs": [ + "add_201" + ], + "output": "_to_copy_264", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_62", + "inputs": [ + "_to_copy_264" + ], + "output": "pow_62", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_61", + "inputs": [ + "pow_62" + ], + "output": "mean_61", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_202", + "inputs": [ + "mean_61" + ], + "output": "add_202", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_91", + "inputs": [ + "add_202" + ], + "output": "rsqrt_91", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_257", + "inputs": [ + "_to_copy_264", + "rsqrt_91" + ], + "output": "mul_257", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_265", + "inputs": [ + "p_model_model_language_model_layers_19_input_layernorm_weight" + ], + "output": "_to_copy_265", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_203", + "inputs": [ + "_to_copy_265" + ], + "output": "add_203", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_258", + "inputs": [ + "mul_257", + "add_203" + ], + "output": "mul_258", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_46", + "inputs": [ + "mul_258" + ], + "output": "type_as_46", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_148", + "inputs": [ + "type_as_46", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight" + ], + "output": "linear_148", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl" + }, + { + "node": "view_125", + "inputs": [ + "linear_148" + ], + "output": "view_125", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl" + }, + { + "node": "slice_170", + "inputs": [ + "view_125" + ], + "output": "slice_170", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c956fa9d3342793cbd859a6e47baea9cfafaac162fc2accf4b8f4728d0f604.wgsl" + }, + { + "node": "slice_171", + "inputs": [ + "view_125" + ], + "output": "slice_171", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl" + }, + { + "node": "clone_31", + "inputs": [ + "slice_171" + ], + "output": "clone_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_unsafe_view_12", + "inputs": [ + "clone_31" + ], + "output": "_unsafe_view_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_126", + "inputs": [ + "slice_170" + ], + "output": "view_126", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_266", + "inputs": [ + "view_126" + ], + "output": "_to_copy_266", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_63", + "inputs": [ + "_to_copy_266" + ], + "output": "pow_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_62", + "inputs": [ + "pow_63" + ], + "output": "mean_62", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl" + }, + { + "node": "add_204", + "inputs": [ + "mean_62" + ], + "output": "add_204", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl" + }, + { + "node": "rsqrt_92", + "inputs": [ + "add_204" + ], + "output": "rsqrt_92", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl" + }, + { + "node": "mul_259", + "inputs": [ + "_to_copy_266", + "rsqrt_92" + ], + "output": "mul_259", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl" + }, + { + "node": "_to_copy_267", + "inputs": [ + "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + ], + "output": "_to_copy_267", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_205", + "inputs": [ + "_to_copy_267" + ], + "output": "add_205", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_260", + "inputs": [ + "mul_259", + "add_205" + ], + "output": "mul_260", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl" + }, + { + "node": "type_as_47", + "inputs": [ + "mul_260" + ], + "output": "type_as_47", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "transpose_78", + "inputs": [ + "type_as_47" + ], + "output": "transpose_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa1c36207dbf73b89a207971c82a8ed3c252e4b5057c44d46f9646ecce99da76.wgsl" + }, + { + "node": "linear_149", + "inputs": [ + "type_as_46", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight" + ], + "output": "linear_149", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_127", + "inputs": [ + "linear_149" + ], + "output": "view_127", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "_to_copy_268", + "inputs": [ + "view_127" + ], + "output": "_to_copy_268", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "pow_64", + "inputs": [ + "_to_copy_268" + ], + "output": "pow_64", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl" + }, + { + "node": "mean_63", + "inputs": [ + "pow_64" + ], + "output": "mean_63", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl" + }, + { + "node": "add_206", + "inputs": [ + "mean_63" + ], + "output": "add_206", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl" + }, + { + "node": "rsqrt_93", + "inputs": [ + "add_206" + ], + "output": "rsqrt_93", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl" + }, + { + "node": "mul_261", + "inputs": [ + "_to_copy_268", + "rsqrt_93" + ], + "output": "mul_261", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl" + }, + { + "node": "_to_copy_269", + "inputs": [ + "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + ], + "output": "_to_copy_269", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_207", + "inputs": [ + "_to_copy_269" + ], + "output": "add_207", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_262", + "inputs": [ + "mul_261", + "add_207" + ], + "output": "mul_262", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl" + }, + { + "node": "type_as_48", + "inputs": [ + "mul_262" + ], + "output": "type_as_48", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f8a4395bd0356791917d7ef0469f9b7d983d33d4b7ad044d66852ead8a86971b.wgsl" + }, + { + "node": "transpose_79", + "inputs": [ + "type_as_48" + ], + "output": "transpose_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "linear_150", + "inputs": [ + "type_as_46", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight" + ], + "output": "linear_150", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_128", + "inputs": [ + "linear_150" + ], + "output": "view_128", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "transpose_80", + "inputs": [ + "view_128" + ], + "output": "transpose_80", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "unsqueeze_36", + "inputs": [ + "_to_copy_1" + ], + "output": "unsqueeze_36", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "unsqueeze_37", + "inputs": [ + "_to_copy_2" + ], + "output": "unsqueeze_37", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "slice_172", + "inputs": [ + "transpose_78" + ], + "output": "slice_172", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl" + }, + { + "node": "slice_173", + "inputs": [ + "transpose_78" + ], + "output": "slice_173", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl" + }, + { + "node": "slice_174", + "inputs": [ + "transpose_79" + ], + "output": "slice_174", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e697b1bc29cd0d08bba0c6a455a7e4b02a3c2e2535304faff2c04b6976617c7c.wgsl" + }, + { + "node": "slice_175", + "inputs": [ + "transpose_79" + ], + "output": "slice_175", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl" + }, + { + "node": "mul_263", + "inputs": [ + "slice_172", + "unsqueeze_36" + ], + "output": "mul_263", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "slice_176", + "inputs": [ + "slice_172" + ], + "output": "slice_176", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl" + }, + { + "node": "slice_177", + "inputs": [ + "slice_172" + ], + "output": "slice_177", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl" + }, + { + "node": "neg_23", + "inputs": [ + "slice_177" + ], + "output": "neg_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl" + }, + { + "node": "cat_32", + "inputs": [ + "neg_23", + "slice_176" + ], + "output": "cat_32", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl" + }, + { + "node": "mul_264", + "inputs": [ + "cat_32", + "unsqueeze_37" + ], + "output": "mul_264", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "add_208", + "inputs": [ + "mul_263", + "mul_264" + ], + "output": "add_208", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl" + }, + { + "node": "mul_265", + "inputs": [ + "slice_174", + "unsqueeze_36" + ], + "output": "mul_265", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "slice_178", + "inputs": [ + "slice_174" + ], + "output": "slice_178", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/fd2084f233c932d1160371a9a522dbc8abdf416285eef54e02e18e4738d12675.wgsl" + }, + { + "node": "slice_179", + "inputs": [ + "slice_174" + ], + "output": "slice_179", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl" + }, + { + "node": "neg_24", + "inputs": [ + "slice_179" + ], + "output": "neg_24", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl" + }, + { + "node": "cat_33", + "inputs": [ + "neg_24", + "slice_178" + ], + "output": "cat_33", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl" + }, + { + "node": "mul_266", + "inputs": [ + "cat_33", + "unsqueeze_37" + ], + "output": "mul_266", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "add_209", + "inputs": [ + "mul_265", + "mul_266" + ], + "output": "add_209", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl" + }, + { + "node": "cat_34", + "inputs": [ + "add_208", + "slice_173" + ], + "output": "cat_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl" + }, + { + "node": "cat_35", + "inputs": [ + "add_209", + "slice_175" + ], + "output": "cat_35", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl" + }, + { + "node": "cache_update_8", + "inputs": [ + "cat_35", + "position_ids" + ], + "output": "cache_update_8", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_38", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "cache_update_9", + "inputs": [ + "transpose_80", + "position_ids" + ], + "output": "cache_update_9", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_39", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "unsqueeze_38", + "inputs": [ + "cache_update_8" + ], + "output": "unsqueeze_38", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_10", + "inputs": [ + "unsqueeze_38" + ], + "output": "expand_10", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_32", + "inputs": [ + "expand_10" + ], + "output": "clone_32", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_13", + "inputs": [ + "clone_32" + ], + "output": "_unsafe_view_13", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "unsqueeze_39", + "inputs": [ + "cache_update_9" + ], + "output": "unsqueeze_39", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_11", + "inputs": [ + "unsqueeze_39" + ], + "output": "expand_11", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_33", + "inputs": [ + "expand_11" + ], + "output": "clone_33", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_14", + "inputs": [ + "clone_33" + ], + "output": "_unsafe_view_14", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "transpose_81", + "inputs": [ + "_unsafe_view_13" + ], + "output": "transpose_81", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl" + }, + { + "node": "matmul_9", + "inputs": [ + "cat_34", + "transpose_81" + ], + "output": "matmul_9", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl" + }, + { + "node": "mul_267", + "inputs": [ + "matmul_9" + ], + "output": "mul_267", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl" + }, + { + "node": "add_210", + "inputs": [ + "mul_267", + "attention_mask" + ], + "output": "add_210", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl" + }, + { + "node": "softmax_4", + "inputs": [ + "add_210" + ], + "output": "softmax_4", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f40727952ccdc71060abbdab6a141b0497a38b727d2f5b943fe826ce90497bd1.wgsl" + }, + { + "node": "_to_copy_270", + "inputs": [ + "softmax_4" + ], + "output": "_to_copy_270", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl" + }, + { + "node": "matmul_10", + "inputs": [ + "_to_copy_270", + "_unsafe_view_14" + ], + "output": "matmul_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl" + }, + { + "node": "transpose_82", + "inputs": [ + "matmul_10" + ], + "output": "transpose_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl" + }, + { + "node": "clone_34", + "inputs": [ + "transpose_82" + ], + "output": "clone_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_129", + "inputs": [ + "clone_34" + ], + "output": "view_129", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_19", + "inputs": [ + "_unsafe_view_12" + ], + "output": "sigmoid_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl" + }, + { + "node": "mul_268", + "inputs": [ + "view_129", + "sigmoid_19" + ], + "output": "mul_268", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/efd4310107ed3db32f1792f1360971ff7ce4234ab64fda03dcaaa9bff52bb912.wgsl" + }, + { + "node": "linear_151", + "inputs": [ + "mul_268", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight" + ], + "output": "linear_151", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_211", + "inputs": [ + "add_201", + "linear_151" + ], + "output": "add_211", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_271", + "inputs": [ + "add_211" + ], + "output": "_to_copy_271", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_65", + "inputs": [ + "_to_copy_271" + ], + "output": "pow_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_64", + "inputs": [ + "pow_65" + ], + "output": "mean_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_212", + "inputs": [ + "mean_64" + ], + "output": "add_212", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_94", + "inputs": [ + "add_212" + ], + "output": "rsqrt_94", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_269", + "inputs": [ + "_to_copy_271", + "rsqrt_94" + ], + "output": "mul_269", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_272", + "inputs": [ + "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + ], + "output": "_to_copy_272", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_213", + "inputs": [ + "_to_copy_272" + ], + "output": "add_213", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_270", + "inputs": [ + "mul_269", + "add_213" + ], + "output": "mul_270", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_49", + "inputs": [ + "mul_270" + ], + "output": "type_as_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_152", + "inputs": [ + "type_as_49", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight" + ], + "output": "linear_152", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_153", + "inputs": [ + "type_as_49", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_271", + "inputs": [ + "linear_152", + "linear_153" + ], + "output": "mul_271", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_154", + "inputs": [ + "mul_271", + "p_model_model_language_model_layers_19_mlp_down_proj_weight" + ], + "output": "linear_154", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_214", + "inputs": [ + "add_211", + "linear_154" + ], + "output": "add_214", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_273", + "inputs": [ + "add_214" + ], + "output": "_to_copy_273", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_66", + "inputs": [ + "_to_copy_273" + ], + "output": "pow_66", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_65", + "inputs": [ + "pow_66" + ], + "output": "mean_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_215", + "inputs": [ + "mean_65" + ], + "output": "add_215", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_95", + "inputs": [ + "add_215" + ], + "output": "rsqrt_95", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_272", + "inputs": [ + "_to_copy_273", + "rsqrt_95" + ], + "output": "mul_272", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_274", + "inputs": [ + "p_model_model_language_model_layers_20_input_layernorm_weight" + ], + "output": "_to_copy_274", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_216", + "inputs": [ + "_to_copy_274" + ], + "output": "add_216", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_273", + "inputs": [ + "mul_272", + "add_216" + ], + "output": "mul_273", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_50", + "inputs": [ + "mul_273" + ], + "output": "type_as_50", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_155", + "inputs": [ + "type_as_50", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_155", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_83", + "inputs": [ + "linear_155" + ], + "output": "transpose_83", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_156", + "inputs": [ + "type_as_50", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight" + ], + "output": "linear_156", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_130", + "inputs": [ + "linear_156" + ], + "output": "view_130", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_157", + "inputs": [ + "type_as_50", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight" + ], + "output": "linear_157", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_158", + "inputs": [ + "type_as_50", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight" + ], + "output": "linear_158", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_36", + "inputs": [ + "states_40", + "transpose_83" + ], + "output": "cat_36", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_180", + "inputs": [ + "cat_36" + ], + "output": "slice_180", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_40" + }, + { + "node": "squeeze_15", + "inputs": [ + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + ], + "output": "squeeze_15", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_40", + "inputs": [ + "squeeze_15" + ], + "output": "unsqueeze_40", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_15", + "inputs": [ + "cat_36", + "unsqueeze_40" + ], + "output": "conv1d_15", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_181", + "inputs": [ + "conv1d_15" + ], + "output": "slice_181", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_50", + "inputs": [ + "slice_181" + ], + "output": "silu_50", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_182", + "inputs": [ + "silu_50" + ], + "output": "slice_182", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_84", + "inputs": [ + "slice_182" + ], + "output": "transpose_84", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_183", + "inputs": [ + "transpose_84" + ], + "output": "slice_183", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_184", + "inputs": [ + "transpose_84" + ], + "output": "slice_184", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_185", + "inputs": [ + "transpose_84" + ], + "output": "slice_185", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_131", + "inputs": [ + "slice_183" + ], + "output": "view_131", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_132", + "inputs": [ + "slice_184" + ], + "output": "view_132", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_133", + "inputs": [ + "slice_185" + ], + "output": "view_133", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_20", + "inputs": [ + "linear_157" + ], + "output": "sigmoid_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_275", + "inputs": [ + "p_model_model_language_model_layers_20_linear_attn_a_log" + ], + "output": "_to_copy_275", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_15", + "inputs": [ + "_to_copy_275" + ], + "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": "_to_copy_276", + "inputs": [ + "linear_158" + ], + "output": "_to_copy_276", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_217", + "inputs": [ + "_to_copy_276", + "p_model_model_language_model_layers_20_linear_attn_dt_bias" + ], + "output": "add_217", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_131" + ], + "output": "mul_275", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_31", + "inputs": [ + "mul_275" + ], + "output": "sum_31", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_218", + "inputs": [ + "sum_31" + ], + "output": "add_218", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_96", + "inputs": [ + "add_218" + ], + "output": "rsqrt_96", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_276", + "inputs": [ + "view_131", + "rsqrt_96" + ], + "output": "mul_276", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_277", + "inputs": [ + "view_132" + ], + "output": "mul_277", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_32", + "inputs": [ + "mul_277" + ], + "output": "sum_32", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_219", + "inputs": [ + "sum_32" + ], + "output": "add_219", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_97", + "inputs": [ + "add_219" + ], + "output": "rsqrt_97", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_278", + "inputs": [ + "view_132", + "rsqrt_97" + ], + "output": "mul_278", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_277", + "inputs": [ + "mul_276" + ], + "output": "_to_copy_277", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_278", + "inputs": [ + "mul_278" + ], + "output": "_to_copy_278", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_279", + "inputs": [ + "view_133" + ], + "output": "_to_copy_279", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_280", + "inputs": [ + "sigmoid_20" + ], + "output": "_to_copy_280", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_15", + "inputs": [ + "_to_copy_277", + "_to_copy_278", + "_to_copy_279", + "mul_274", + "_to_copy_280", + "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": "clone_35", + "inputs": [ + "transpose_85" + ], + "output": "clone_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_281", + "inputs": [ + "clone_35" + ], + "output": "_to_copy_281", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_134", + "inputs": [ + "_to_copy_281" + ], + "output": "view_134", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_135", + "inputs": [ + "view_130" + ], + "output": "view_135", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_282", + "inputs": [ + "view_134" + ], + "output": "_to_copy_282", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_67", + "inputs": [ + "_to_copy_282" + ], + "output": "pow_67", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_66", + "inputs": [ + "pow_67" + ], + "output": "mean_66", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_220", + "inputs": [ + "mean_66" + ], + "output": "add_220", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_98", + "inputs": [ + "add_220" + ], + "output": "rsqrt_98", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_279", + "inputs": [ + "_to_copy_282", + "rsqrt_98" + ], + "output": "mul_279", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_283", + "inputs": [ + "mul_279" + ], + "output": "_to_copy_283", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_280", + "inputs": [ + "p_model_model_language_model_layers_20_linear_attn_norm_weight", + "_to_copy_283" + ], + "output": "mul_280", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_284", + "inputs": [ + "view_135" + ], + "output": "_to_copy_284", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_281", + "inputs": [ + "_to_copy_284", + "mul_280" + ], + "output": "mul_281", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_285", + "inputs": [ + "mul_281" + ], + "output": "_to_copy_285", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_136", + "inputs": [ + "_to_copy_285" + ], + "output": "view_136", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_159", + "inputs": [ + "view_136", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight" + ], + "output": "linear_159", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_221", + "inputs": [ + "add_214", + "linear_159" + ], + "output": "add_221", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_286", + "inputs": [ + "add_221" + ], + "output": "_to_copy_286", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_68", + "inputs": [ + "_to_copy_286" + ], + "output": "pow_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_67", + "inputs": [ + "pow_68" + ], + "output": "mean_67", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_222", + "inputs": [ + "mean_67" + ], + "output": "add_222", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_99", + "inputs": [ + "add_222" + ], + "output": "rsqrt_99", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_282", + "inputs": [ + "_to_copy_286", + "rsqrt_99" + ], + "output": "mul_282", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_287", + "inputs": [ + "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + ], + "output": "_to_copy_287", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_223", + "inputs": [ + "_to_copy_287" + ], + "output": "add_223", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_283", + "inputs": [ + "mul_282", + "add_223" + ], + "output": "mul_283", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_51", + "inputs": [ + "mul_283" + ], + "output": "type_as_51", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_160", + "inputs": [ + "type_as_51", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight" + ], + "output": "linear_160", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_161", + "inputs": [ + "type_as_51", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_284", + "inputs": [ + "linear_160", + "linear_161" + ], + "output": "mul_284", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_162", + "inputs": [ + "mul_284", + "p_model_model_language_model_layers_20_mlp_down_proj_weight" + ], + "output": "linear_162", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_224", + "inputs": [ + "add_221", + "linear_162" + ], + "output": "add_224", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_288", + "inputs": [ + "add_224" + ], + "output": "_to_copy_288", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_69", + "inputs": [ + "_to_copy_288" + ], + "output": "pow_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_68", + "inputs": [ + "pow_69" + ], + "output": "mean_68", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_225", + "inputs": [ + "mean_68" + ], + "output": "add_225", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_100", + "inputs": [ + "add_225" + ], + "output": "rsqrt_100", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_285", + "inputs": [ + "_to_copy_288", + "rsqrt_100" + ], + "output": "mul_285", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_289", + "inputs": [ + "p_model_model_language_model_layers_21_input_layernorm_weight" + ], + "output": "_to_copy_289", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_226", + "inputs": [ + "_to_copy_289" + ], + "output": "add_226", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_286", + "inputs": [ + "mul_285", + "add_226" + ], + "output": "mul_286", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_52", + "inputs": [ + "mul_286" + ], + "output": "type_as_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_163", + "inputs": [ + "type_as_52", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_163", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_86", + "inputs": [ + "linear_163" + ], + "output": "transpose_86", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_164", + "inputs": [ + "type_as_52", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight" + ], + "output": "linear_164", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_137", + "inputs": [ + "linear_164" + ], + "output": "view_137", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_165", + "inputs": [ + "type_as_52", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight" + ], + "output": "linear_165", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_166", + "inputs": [ + "type_as_52", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight" + ], + "output": "linear_166", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_37", + "inputs": [ + "states_42", + "transpose_86" + ], + "output": "cat_37", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_188", + "inputs": [ + "cat_37" + ], + "output": "slice_188", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_42" + }, + { + "node": "squeeze_16", + "inputs": [ + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + ], + "output": "squeeze_16", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_41", + "inputs": [ + "squeeze_16" + ], + "output": "unsqueeze_41", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_16", + "inputs": [ + "cat_37", + "unsqueeze_41" + ], + "output": "conv1d_16", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_189", + "inputs": [ + "conv1d_16" + ], + "output": "slice_189", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_53", + "inputs": [ + "slice_189" + ], + "output": "silu_53", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_190", + "inputs": [ + "silu_53" + ], + "output": "slice_190", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_87", + "inputs": [ + "slice_190" + ], + "output": "transpose_87", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_191", + "inputs": [ + "transpose_87" + ], + "output": "slice_191", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_192", + "inputs": [ + "transpose_87" + ], + "output": "slice_192", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_193", + "inputs": [ + "transpose_87" + ], + "output": "slice_193", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_138", + "inputs": [ + "slice_191" + ], + "output": "view_138", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_139", + "inputs": [ + "slice_192" + ], + "output": "view_139", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_140", + "inputs": [ + "slice_193" + ], + "output": "view_140", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_21", + "inputs": [ + "linear_165" + ], + "output": "sigmoid_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_290", + "inputs": [ + "p_model_model_language_model_layers_21_linear_attn_a_log" + ], + "output": "_to_copy_290", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_16", + "inputs": [ + "_to_copy_290" + ], + "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": "_to_copy_291", + "inputs": [ + "linear_166" + ], + "output": "_to_copy_291", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_227", + "inputs": [ + "_to_copy_291", + "p_model_model_language_model_layers_21_linear_attn_dt_bias" + ], + "output": "add_227", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_138" + ], + "output": "mul_288", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_33", + "inputs": [ + "mul_288" + ], + "output": "sum_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_228", + "inputs": [ + "sum_33" + ], + "output": "add_228", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_101", + "inputs": [ + "add_228" + ], + "output": "rsqrt_101", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_289", + "inputs": [ + "view_138", + "rsqrt_101" + ], + "output": "mul_289", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_290", + "inputs": [ + "view_139" + ], + "output": "mul_290", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_34", + "inputs": [ + "mul_290" + ], + "output": "sum_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_229", + "inputs": [ + "sum_34" + ], + "output": "add_229", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_102", + "inputs": [ + "add_229" + ], + "output": "rsqrt_102", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_291", + "inputs": [ + "view_139", + "rsqrt_102" + ], + "output": "mul_291", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_292", + "inputs": [ + "mul_289" + ], + "output": "_to_copy_292", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_293", + "inputs": [ + "mul_291" + ], + "output": "_to_copy_293", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_294", + "inputs": [ + "view_140" + ], + "output": "_to_copy_294", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_295", + "inputs": [ + "sigmoid_21" + ], + "output": "_to_copy_295", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_16", + "inputs": [ + "_to_copy_292", + "_to_copy_293", + "_to_copy_294", + "mul_287", + "_to_copy_295", + "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": "clone_36", + "inputs": [ + "transpose_88" + ], + "output": "clone_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_296", + "inputs": [ + "clone_36" + ], + "output": "_to_copy_296", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_141", + "inputs": [ + "_to_copy_296" + ], + "output": "view_141", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_142", + "inputs": [ + "view_137" + ], + "output": "view_142", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_297", + "inputs": [ + "view_141" + ], + "output": "_to_copy_297", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_70", + "inputs": [ + "_to_copy_297" + ], + "output": "pow_70", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_69", + "inputs": [ + "pow_70" + ], + "output": "mean_69", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_230", + "inputs": [ + "mean_69" + ], + "output": "add_230", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_103", + "inputs": [ + "add_230" + ], + "output": "rsqrt_103", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_292", + "inputs": [ + "_to_copy_297", + "rsqrt_103" + ], + "output": "mul_292", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_298", + "inputs": [ + "mul_292" + ], + "output": "_to_copy_298", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_293", + "inputs": [ + "p_model_model_language_model_layers_21_linear_attn_norm_weight", + "_to_copy_298" + ], + "output": "mul_293", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_299", + "inputs": [ + "view_142" + ], + "output": "_to_copy_299", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_294", + "inputs": [ + "_to_copy_299", + "mul_293" + ], + "output": "mul_294", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_300", + "inputs": [ + "mul_294" + ], + "output": "_to_copy_300", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_143", + "inputs": [ + "_to_copy_300" + ], + "output": "view_143", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_167", + "inputs": [ + "view_143", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight" + ], + "output": "linear_167", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_231", + "inputs": [ + "add_224", + "linear_167" + ], + "output": "add_231", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_301", + "inputs": [ + "add_231" + ], + "output": "_to_copy_301", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_71", + "inputs": [ + "_to_copy_301" + ], + "output": "pow_71", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_70", + "inputs": [ + "pow_71" + ], + "output": "mean_70", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_232", + "inputs": [ + "mean_70" + ], + "output": "add_232", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_104", + "inputs": [ + "add_232" + ], + "output": "rsqrt_104", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_295", + "inputs": [ + "_to_copy_301", + "rsqrt_104" + ], + "output": "mul_295", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_302", + "inputs": [ + "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + ], + "output": "_to_copy_302", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_233", + "inputs": [ + "_to_copy_302" + ], + "output": "add_233", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_296", + "inputs": [ + "mul_295", + "add_233" + ], + "output": "mul_296", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_53", + "inputs": [ + "mul_296" + ], + "output": "type_as_53", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_168", + "inputs": [ + "type_as_53", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight" + ], + "output": "linear_168", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_169", + "inputs": [ + "type_as_53", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_297", + "inputs": [ + "linear_168", + "linear_169" + ], + "output": "mul_297", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_170", + "inputs": [ + "mul_297", + "p_model_model_language_model_layers_21_mlp_down_proj_weight" + ], + "output": "linear_170", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_234", + "inputs": [ + "add_231", + "linear_170" + ], + "output": "add_234", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_303", + "inputs": [ + "add_234" + ], + "output": "_to_copy_303", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_72", + "inputs": [ + "_to_copy_303" + ], + "output": "pow_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_71", + "inputs": [ + "pow_72" + ], + "output": "mean_71", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_235", + "inputs": [ + "mean_71" + ], + "output": "add_235", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_105", + "inputs": [ + "add_235" + ], + "output": "rsqrt_105", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_298", + "inputs": [ + "_to_copy_303", + "rsqrt_105" + ], + "output": "mul_298", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_304", + "inputs": [ + "p_model_model_language_model_layers_22_input_layernorm_weight" + ], + "output": "_to_copy_304", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_236", + "inputs": [ + "_to_copy_304" + ], + "output": "add_236", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_299", + "inputs": [ + "mul_298", + "add_236" + ], + "output": "mul_299", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_54", + "inputs": [ + "mul_299" + ], + "output": "type_as_54", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_171", + "inputs": [ + "type_as_54", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight" + ], + "output": "linear_171", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "transpose_89", + "inputs": [ + "linear_171" + ], + "output": "transpose_89", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl" + }, + { + "node": "linear_172", + "inputs": [ + "type_as_54", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight" + ], + "output": "linear_172", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "view_144", + "inputs": [ + "linear_172" + ], + "output": "view_144", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_173", + "inputs": [ + "type_as_54", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight" + ], + "output": "linear_173", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "linear_174", + "inputs": [ + "type_as_54", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight" + ], + "output": "linear_174", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl" + }, + { + "node": "cat_38", + "inputs": [ + "states_44", + "transpose_89" + ], + "output": "cat_38", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl" + }, + { + "node": "slice_196", + "inputs": [ + "cat_38" + ], + "output": "slice_196", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl", + "inPlaceOf": "states_44" + }, + { + "node": "squeeze_17", + "inputs": [ + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + ], + "output": "squeeze_17", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "unsqueeze_42", + "inputs": [ + "squeeze_17" + ], + "output": "unsqueeze_42", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl" + }, + { + "node": "conv1d_17", + "inputs": [ + "cat_38", + "unsqueeze_42" + ], + "output": "conv1d_17", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl" + }, + { + "node": "slice_197", + "inputs": [ + "conv1d_17" + ], + "output": "slice_197", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl" + }, + { + "node": "silu_56", + "inputs": [ + "slice_197" + ], + "output": "silu_56", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl" + }, + { + "node": "slice_198", + "inputs": [ + "silu_56" + ], + "output": "slice_198", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl" + }, + { + "node": "transpose_90", + "inputs": [ + "slice_198" + ], + "output": "transpose_90", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl" + }, + { + "node": "slice_199", + "inputs": [ + "transpose_90" + ], + "output": "slice_199", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl" + }, + { + "node": "slice_200", + "inputs": [ + "transpose_90" + ], + "output": "slice_200", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl" + }, + { + "node": "slice_201", + "inputs": [ + "transpose_90" + ], + "output": "slice_201", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl" + }, + { + "node": "view_145", + "inputs": [ + "slice_199" + ], + "output": "view_145", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_146", + "inputs": [ + "slice_200" + ], + "output": "view_146", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_147", + "inputs": [ + "slice_201" + ], + "output": "view_147", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_22", + "inputs": [ + "linear_173" + ], + "output": "sigmoid_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl" + }, + { + "node": "_to_copy_305", + "inputs": [ + "p_model_model_language_model_layers_22_linear_attn_a_log" + ], + "output": "_to_copy_305", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl" + }, + { + "node": "exp_17", + "inputs": [ + "_to_copy_305" + ], + "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": "_to_copy_306", + "inputs": [ + "linear_174" + ], + "output": "_to_copy_306", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "add_237", + "inputs": [ + "_to_copy_306", + "p_model_model_language_model_layers_22_linear_attn_dt_bias" + ], + "output": "add_237", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.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": [ + "view_145" + ], + "output": "mul_301", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_35", + "inputs": [ + "mul_301" + ], + "output": "sum_35", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_238", + "inputs": [ + "sum_35" + ], + "output": "add_238", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_106", + "inputs": [ + "add_238" + ], + "output": "rsqrt_106", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_302", + "inputs": [ + "view_145", + "rsqrt_106" + ], + "output": "mul_302", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "mul_303", + "inputs": [ + "view_146" + ], + "output": "mul_303", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl" + }, + { + "node": "sum_36", + "inputs": [ + "mul_303" + ], + "output": "sum_36", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl" + }, + { + "node": "add_239", + "inputs": [ + "sum_36" + ], + "output": "add_239", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl" + }, + { + "node": "rsqrt_107", + "inputs": [ + "add_239" + ], + "output": "rsqrt_107", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl" + }, + { + "node": "mul_304", + "inputs": [ + "view_146", + "rsqrt_107" + ], + "output": "mul_304", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl" + }, + { + "node": "_to_copy_307", + "inputs": [ + "mul_302" + ], + "output": "_to_copy_307", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_308", + "inputs": [ + "mul_304" + ], + "output": "_to_copy_308", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_309", + "inputs": [ + "view_147" + ], + "output": "_to_copy_309", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "_to_copy_310", + "inputs": [ + "sigmoid_22" + ], + "output": "_to_copy_310", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl" + }, + { + "node": "normalized_gated_delta_rule_17", + "inputs": [ + "_to_copy_307", + "_to_copy_308", + "_to_copy_309", + "mul_300", + "_to_copy_310", + "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": "clone_37", + "inputs": [ + "transpose_91" + ], + "output": "clone_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl" + }, + { + "node": "_to_copy_311", + "inputs": [ + "clone_37" + ], + "output": "_to_copy_311", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_148", + "inputs": [ + "_to_copy_311" + ], + "output": "view_148", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_149", + "inputs": [ + "view_144" + ], + "output": "view_149", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_312", + "inputs": [ + "view_148" + ], + "output": "_to_copy_312", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_73", + "inputs": [ + "_to_copy_312" + ], + "output": "pow_73", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_72", + "inputs": [ + "pow_73" + ], + "output": "mean_72", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl" + }, + { + "node": "add_240", + "inputs": [ + "mean_72" + ], + "output": "add_240", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_108", + "inputs": [ + "add_240" + ], + "output": "rsqrt_108", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_305", + "inputs": [ + "_to_copy_312", + "rsqrt_108" + ], + "output": "mul_305", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl" + }, + { + "node": "_to_copy_313", + "inputs": [ + "mul_305" + ], + "output": "_to_copy_313", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "mul_306", + "inputs": [ + "p_model_model_language_model_layers_22_linear_attn_norm_weight", + "_to_copy_313" + ], + "output": "mul_306", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl" + }, + { + "node": "_to_copy_314", + "inputs": [ + "view_149" + ], + "output": "_to_copy_314", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "mul_307", + "inputs": [ + "_to_copy_314", + "mul_306" + ], + "output": "mul_307", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl" + }, + { + "node": "_to_copy_315", + "inputs": [ + "mul_307" + ], + "output": "_to_copy_315", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "view_150", + "inputs": [ + "_to_copy_315" + ], + "output": "view_150", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_175", + "inputs": [ + "view_150", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight" + ], + "output": "linear_175", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_241", + "inputs": [ + "add_234", + "linear_175" + ], + "output": "add_241", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_316", + "inputs": [ + "add_241" + ], + "output": "_to_copy_316", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_74", + "inputs": [ + "_to_copy_316" + ], + "output": "pow_74", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_73", + "inputs": [ + "pow_74" + ], + "output": "mean_73", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_242", + "inputs": [ + "mean_73" + ], + "output": "add_242", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_109", + "inputs": [ + "add_242" + ], + "output": "rsqrt_109", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_308", + "inputs": [ + "_to_copy_316", + "rsqrt_109" + ], + "output": "mul_308", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_317", + "inputs": [ + "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + ], + "output": "_to_copy_317", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_243", + "inputs": [ + "_to_copy_317" + ], + "output": "add_243", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_309", + "inputs": [ + "mul_308", + "add_243" + ], + "output": "mul_309", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_55", + "inputs": [ + "mul_309" + ], + "output": "type_as_55", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_176", + "inputs": [ + "type_as_55", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight" + ], + "output": "linear_176", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_177", + "inputs": [ + "type_as_55", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_310", + "inputs": [ + "linear_176", + "linear_177" + ], + "output": "mul_310", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_178", + "inputs": [ + "mul_310", + "p_model_model_language_model_layers_22_mlp_down_proj_weight" + ], + "output": "linear_178", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_244", + "inputs": [ + "add_241", + "linear_178" + ], + "output": "add_244", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_318", + "inputs": [ + "add_244" + ], + "output": "_to_copy_318", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_75", + "inputs": [ + "_to_copy_318" + ], + "output": "pow_75", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_74", + "inputs": [ + "pow_75" + ], + "output": "mean_74", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_245", + "inputs": [ + "mean_74" + ], + "output": "add_245", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_110", + "inputs": [ + "add_245" + ], + "output": "rsqrt_110", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_311", + "inputs": [ + "_to_copy_318", + "rsqrt_110" + ], + "output": "mul_311", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_319", + "inputs": [ + "p_model_model_language_model_layers_23_input_layernorm_weight" + ], + "output": "_to_copy_319", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_246", + "inputs": [ + "_to_copy_319" + ], + "output": "add_246", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_312", + "inputs": [ + "mul_311", + "add_246" + ], + "output": "mul_312", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_56", + "inputs": [ + "mul_312" + ], + "output": "type_as_56", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_179", + "inputs": [ + "type_as_56", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight" + ], + "output": "linear_179", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl" + }, + { + "node": "view_151", + "inputs": [ + "linear_179" + ], + "output": "view_151", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl" + }, + { + "node": "slice_204", + "inputs": [ + "view_151" + ], + "output": "slice_204", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c956fa9d3342793cbd859a6e47baea9cfafaac162fc2accf4b8f4728d0f604.wgsl" + }, + { + "node": "slice_205", + "inputs": [ + "view_151" + ], + "output": "slice_205", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl" + }, + { + "node": "clone_38", + "inputs": [ + "slice_205" + ], + "output": "clone_38", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_unsafe_view_15", + "inputs": [ + "clone_38" + ], + "output": "_unsafe_view_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_152", + "inputs": [ + "slice_204" + ], + "output": "view_152", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "_to_copy_320", + "inputs": [ + "view_152" + ], + "output": "_to_copy_320", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_76", + "inputs": [ + "_to_copy_320" + ], + "output": "pow_76", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_75", + "inputs": [ + "pow_76" + ], + "output": "mean_75", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl" + }, + { + "node": "add_247", + "inputs": [ + "mean_75" + ], + "output": "add_247", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl" + }, + { + "node": "rsqrt_111", + "inputs": [ + "add_247" + ], + "output": "rsqrt_111", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl" + }, + { + "node": "mul_313", + "inputs": [ + "_to_copy_320", + "rsqrt_111" + ], + "output": "mul_313", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl" + }, + { + "node": "_to_copy_321", + "inputs": [ + "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + ], + "output": "_to_copy_321", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_248", + "inputs": [ + "_to_copy_321" + ], + "output": "add_248", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_314", + "inputs": [ + "mul_313", + "add_248" + ], + "output": "mul_314", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl" + }, + { + "node": "type_as_57", + "inputs": [ + "mul_314" + ], + "output": "type_as_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "transpose_92", + "inputs": [ + "type_as_57" + ], + "output": "transpose_92", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa1c36207dbf73b89a207971c82a8ed3c252e4b5057c44d46f9646ecce99da76.wgsl" + }, + { + "node": "linear_180", + "inputs": [ + "type_as_56", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight" + ], + "output": "linear_180", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_153", + "inputs": [ + "linear_180" + ], + "output": "view_153", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "_to_copy_322", + "inputs": [ + "view_153" + ], + "output": "_to_copy_322", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "pow_77", + "inputs": [ + "_to_copy_322" + ], + "output": "pow_77", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl" + }, + { + "node": "mean_76", + "inputs": [ + "pow_77" + ], + "output": "mean_76", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl" + }, + { + "node": "add_249", + "inputs": [ + "mean_76" + ], + "output": "add_249", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl" + }, + { + "node": "rsqrt_112", + "inputs": [ + "add_249" + ], + "output": "rsqrt_112", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl" + }, + { + "node": "mul_315", + "inputs": [ + "_to_copy_322", + "rsqrt_112" + ], + "output": "mul_315", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl" + }, + { + "node": "_to_copy_323", + "inputs": [ + "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + ], + "output": "_to_copy_323", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/f545db742864db922611db9427c0784c9ced6fe4fb27a75b15cee7f83a04fc7a.wgsl" + }, + { + "node": "add_250", + "inputs": [ + "_to_copy_323" + ], + "output": "add_250", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl" + }, + { + "node": "mul_316", + "inputs": [ + "mul_315", + "add_250" + ], + "output": "mul_316", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl" + }, + { + "node": "type_as_58", + "inputs": [ + "mul_316" + ], + "output": "type_as_58", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f8a4395bd0356791917d7ef0469f9b7d983d33d4b7ad044d66852ead8a86971b.wgsl" + }, + { + "node": "transpose_93", + "inputs": [ + "type_as_58" + ], + "output": "transpose_93", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "linear_181", + "inputs": [ + "type_as_56", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight" + ], + "output": "linear_181", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl" + }, + { + "node": "view_154", + "inputs": [ + "linear_181" + ], + "output": "view_154", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl" + }, + { + "node": "transpose_94", + "inputs": [ + "view_154" + ], + "output": "transpose_94", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f76e817d0fd4ad9c1f1eb8f983fa75d36bf001ffeebe8cc021053ab41adba73e.wgsl" + }, + { + "node": "unsqueeze_43", + "inputs": [ + "_to_copy_1" + ], + "output": "unsqueeze_43", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "unsqueeze_44", + "inputs": [ + "_to_copy_2" + ], + "output": "unsqueeze_44", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl" + }, + { + "node": "slice_206", + "inputs": [ + "transpose_92" + ], + "output": "slice_206", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl" + }, + { + "node": "slice_207", + "inputs": [ + "transpose_92" + ], + "output": "slice_207", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl" + }, + { + "node": "slice_208", + "inputs": [ + "transpose_93" + ], + "output": "slice_208", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e697b1bc29cd0d08bba0c6a455a7e4b02a3c2e2535304faff2c04b6976617c7c.wgsl" + }, + { + "node": "slice_209", + "inputs": [ + "transpose_93" + ], + "output": "slice_209", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl" + }, + { + "node": "mul_317", + "inputs": [ + "slice_206", + "unsqueeze_43" + ], + "output": "mul_317", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "slice_210", + "inputs": [ + "slice_206" + ], + "output": "slice_210", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl" + }, + { + "node": "slice_211", + "inputs": [ + "slice_206" + ], + "output": "slice_211", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl" + }, + { + "node": "neg_28", + "inputs": [ + "slice_211" + ], + "output": "neg_28", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl" + }, + { + "node": "cat_39", + "inputs": [ + "neg_28", + "slice_210" + ], + "output": "cat_39", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl" + }, + { + "node": "mul_318", + "inputs": [ + "cat_39", + "unsqueeze_44" + ], + "output": "mul_318", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl" + }, + { + "node": "add_251", + "inputs": [ + "mul_317", + "mul_318" + ], + "output": "add_251", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl" + }, + { + "node": "mul_319", + "inputs": [ + "slice_208", + "unsqueeze_43" + ], + "output": "mul_319", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "slice_212", + "inputs": [ + "slice_208" + ], + "output": "slice_212", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/fd2084f233c932d1160371a9a522dbc8abdf416285eef54e02e18e4738d12675.wgsl" + }, + { + "node": "slice_213", + "inputs": [ + "slice_208" + ], + "output": "slice_213", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl" + }, + { + "node": "neg_29", + "inputs": [ + "slice_213" + ], + "output": "neg_29", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl" + }, + { + "node": "cat_40", + "inputs": [ + "neg_29", + "slice_212" + ], + "output": "cat_40", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl" + }, + { + "node": "mul_320", + "inputs": [ + "cat_40", + "unsqueeze_44" + ], + "output": "mul_320", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/e3db4be22be4d51e95cce7d12cb5132ee8ddd18929abc0090ea9522eab965393.wgsl" + }, + { + "node": "add_252", + "inputs": [ + "mul_319", + "mul_320" + ], + "output": "add_252", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl" + }, + { + "node": "cat_41", + "inputs": [ + "add_251", + "slice_207" + ], + "output": "cat_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl" + }, + { + "node": "cat_42", + "inputs": [ + "add_252", + "slice_209" + ], + "output": "cat_42", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl" + }, + { + "node": "cache_update_10", + "inputs": [ + "cat_42", + "position_ids" + ], + "output": "cache_update_10", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_46", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "cache_update_11", + "inputs": [ + "transpose_94", + "position_ids" + ], + "output": "cache_update_11", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_47", + "kernel": "kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl" + }, + { + "node": "unsqueeze_45", + "inputs": [ + "cache_update_10" + ], + "output": "unsqueeze_45", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_12", + "inputs": [ + "unsqueeze_45" + ], + "output": "expand_12", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_39", + "inputs": [ + "expand_12" + ], + "output": "clone_39", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_16", + "inputs": [ + "clone_39" + ], + "output": "_unsafe_view_16", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "unsqueeze_46", + "inputs": [ + "cache_update_11" + ], + "output": "unsqueeze_46", + "workgroups": [ + 32768, + 1, + 1 + ], + "kernel": "kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl" + }, + { + "node": "expand_13", + "inputs": [ + "unsqueeze_46" + ], + "output": "expand_13", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl" + }, + { + "node": "clone_40", + "inputs": [ + "expand_13" + ], + "output": "clone_40", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "_unsafe_view_17", + "inputs": [ + "clone_40" + ], + "output": "_unsafe_view_17", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl" + }, + { + "node": "transpose_95", + "inputs": [ + "_unsafe_view_16" + ], + "output": "transpose_95", + "workgroups": [ + 65535, + 3, + 1 + ], + "kernel": "kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl" + }, + { + "node": "matmul_11", + "inputs": [ + "cat_41", + "transpose_95" + ], + "output": "matmul_11", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl" + }, + { + "node": "mul_321", + "inputs": [ + "matmul_11" + ], + "output": "mul_321", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl" + }, + { + "node": "add_253", + "inputs": [ + "mul_321", + "attention_mask" + ], + "output": "add_253", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl" + }, + { + "node": "softmax_5", + "inputs": [ + "add_253" + ], + "output": "softmax_5", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f40727952ccdc71060abbdab6a141b0497a38b727d2f5b943fe826ce90497bd1.wgsl" + }, + { + "node": "_to_copy_324", + "inputs": [ + "softmax_5" + ], + "output": "_to_copy_324", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl" + }, + { + "node": "matmul_12", + "inputs": [ + "_to_copy_324", + "_unsafe_view_17" + ], + "output": "matmul_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl" + }, + { + "node": "transpose_96", + "inputs": [ + "matmul_12" + ], + "output": "transpose_96", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl" + }, + { + "node": "clone_41", + "inputs": [ + "transpose_96" + ], + "output": "clone_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "view_155", + "inputs": [ + "clone_41" + ], + "output": "view_155", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "sigmoid_23", + "inputs": [ + "_unsafe_view_15" + ], + "output": "sigmoid_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl" + }, + { + "node": "mul_322", + "inputs": [ + "view_155", + "sigmoid_23" + ], + "output": "mul_322", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/efd4310107ed3db32f1792f1360971ff7ce4234ab64fda03dcaaa9bff52bb912.wgsl" + }, + { + "node": "linear_182", + "inputs": [ + "mul_322", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight" + ], + "output": "linear_182", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl" + }, + { + "node": "add_254", + "inputs": [ + "add_244", + "linear_182" + ], + "output": "add_254", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_325", + "inputs": [ + "add_254" + ], + "output": "_to_copy_325", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_78", + "inputs": [ + "_to_copy_325" + ], + "output": "pow_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_77", + "inputs": [ + "pow_78" + ], + "output": "mean_77", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_255", + "inputs": [ + "mean_77" + ], + "output": "add_255", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_113", + "inputs": [ + "add_255" + ], + "output": "rsqrt_113", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_323", + "inputs": [ + "_to_copy_325", + "rsqrt_113" + ], + "output": "mul_323", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_326", + "inputs": [ + "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + ], + "output": "_to_copy_326", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_256", + "inputs": [ + "_to_copy_326" + ], + "output": "add_256", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_324", + "inputs": [ + "mul_323", + "add_256" + ], + "output": "mul_324", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_59", + "inputs": [ + "mul_324" + ], + "output": "type_as_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "linear_183", + "inputs": [ + "type_as_59", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight" + ], + "output": "linear_183", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl" + }, + { + "node": "linear_184", + "inputs": [ + "type_as_59", + "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": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl" + }, + { + "node": "mul_325", + "inputs": [ + "linear_183", + "linear_184" + ], + "output": "mul_325", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/f3defbac8d19992cfa348ec80967aec0c899cc3844ff5d320f19514492df5d9d.wgsl" + }, + { + "node": "linear_185", + "inputs": [ + "mul_325", + "p_model_model_language_model_layers_23_mlp_down_proj_weight" + ], + "output": "linear_185", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl" + }, + { + "node": "add_257", + "inputs": [ + "add_254", + "linear_185" + ], + "output": "add_257", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl" + }, + { + "node": "_to_copy_327", + "inputs": [ + "add_257" + ], + "output": "_to_copy_327", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl" + }, + { + "node": "pow_79", + "inputs": [ + "_to_copy_327" + ], + "output": "pow_79", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f6c4d4177cf7f8a388a006af894d865388fb805d66339c92a711bee10a92e864.wgsl" + }, + { + "node": "mean_78", + "inputs": [ + "pow_79" + ], + "output": "mean_78", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl" + }, + { + "node": "add_258", + "inputs": [ + "mean_78" + ], + "output": "add_258", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl" + }, + { + "node": "rsqrt_114", + "inputs": [ + "add_258" + ], + "output": "rsqrt_114", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl" + }, + { + "node": "mul_326", + "inputs": [ + "_to_copy_327", + "rsqrt_114" + ], + "output": "mul_326", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f8f254ef99c96d854e32eb1471a76fc90ecf7430140c9dbe819450f89a6c1f86.wgsl" + }, + { + "node": "_to_copy_328", + "inputs": [ + "p_model_model_language_model_norm_weight" + ], + "output": "_to_copy_328", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl" + }, + { + "node": "add_259", + "inputs": [ + "_to_copy_328" + ], + "output": "add_259", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl" + }, + { + "node": "mul_327", + "inputs": [ + "mul_326", + "add_259" + ], + "output": "mul_327", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl" + }, + { + "node": "type_as_60", + "inputs": [ + "mul_327" + ], + "output": "type_as_60", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl" + }, + { + "node": "alias", + "inputs": [ + "type_as_60" + ], + "output": "alias", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl" + }, + { + "node": "linear_186__part_0", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_0_32768" + ], + "output": "linear_186__part_0", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_1", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_32768_32768" + ], + "output": "linear_186__part_1", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_2", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_65536_32768" + ], + "output": "linear_186__part_2", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_3", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_98304_32768" + ], + "output": "linear_186__part_3", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_4", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_131072_32768" + ], + "output": "linear_186__part_4", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_5", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_163840_32768" + ], + "output": "linear_186__part_5", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_6", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_196608_32768" + ], + "output": "linear_186__part_6", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl" + }, + { + "node": "linear_186__part_7", + "inputs": [ + "alias", + "p_model_model_language_model_embed_tokens_weight__rows_229376_18944" + ], + "output": "linear_186__part_7", + "workgroups": [ + 2368, + 1, + 1 + ], + "kernel": "kernels/8d9b846ecd9fe2c3acde76737571459a46136953e170abfca48b642fe9dce367.wgsl" + }, + { + "node": "linear_186__cat_0_0", + "inputs": [ + "linear_186__part_0", + "linear_186__part_1", + "linear_186__part_2", + "linear_186__part_3", + "linear_186__part_4", + "linear_186__part_5", + "linear_186__part_6" + ], + "output": "linear_186__cat_0_0", + "workgroups": [ + 14336, + 1, + 1 + ], + "kernel": "kernels/7e90b8d0c7b998d4729541adc2de35c00fb0780d7ca3e89c66cb32638c3b533d.wgsl" + }, + { + "node": "linear_186__cat_0_7", + "inputs": [ + "linear_186__part_7" + ], + "output": "linear_186__cat_0_7", + "workgroups": [ + 1184, + 1, + 1 + ], + "kernel": "kernels/570f05a5f9f892e7a8075cde8b7bee28bc921f6054167eec4b85797f15b873da.wgsl" + }, + { + "node": "linear_186", + "inputs": [ + "linear_186__cat_0_0", + "linear_186__cat_0_7" + ], + "output": "linear_186", + "workgroups": [ + 15520, + 1, + 1 + ], + "kernel": "kernels/ba2b60fc7e0bf8f703260ce0eb714460e8411fe09096a0d1950298537a0b0d3a.wgsl" + } + ], + "memory": { + "slots": [ + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 16777216, + 512, + 2097152, + 512, + 2097152, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 16777216, + 512, + 2097152, + 512, + 2097152, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 16777216, + 512, + 2097152, + 512, + 2097152, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 16777216, + 512, + 2097152, + 512, + 2097152, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 16777216, + 512, + 2097152, + 512, + 2097152, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 32, + 32, + 25165824, + 8388608, + 65536, + 65536, + 49152, + 256, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 16777216, + 512, + 2097152, + 512, + 2097152, + 8388608, + 4096, + 25165824, + 6291456, + 25165824, + 4096, + 128, + 16, + 16, + 32768, + 49152, + 1048576, + 49152, + 1048576, + 49152, + 1048576, + 4194304, + 4194304, + 49152, + 1048576, + 49152, + 1048576, + 49152, + 1048576, + 4194304, + 4194304, + 49152, + 1048576, + 49152, + 1048576, + 49152, + 1048576, + 4194304, + 4194304, + 49152, + 1048576, + 49152, + 1048576, + 49152, + 1048576, + 4194304, + 4194304, + 49152, + 1048576, + 49152, + 1048576, + 49152, + 1048576, + 4194304, + 4194304, + 49152, + 1048576, + 49152, + 1048576, + 49152, + 1048576, + 4194304, + 4194304, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 6291456, + 134217728, + 134217728, + 134217728, + 134217728, + 134217728, + 134217728, + 134217728, + 77594624, + 16384, + 16384, + 16384, + 16384, + 16384, + 16384, + 16384, + 16384, + 16384, + 32768, + 32768, + 49152, + 49152, + 98304, + 135168, + 1081344, + 4194304, + 16777216, + 16777216, + 16777216, + 262144, + 262144, + 151552, + 1835008 + ], + "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_up_proj_weight__int4_scales": 371, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": 372, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": 373, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": 374, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": 375, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": 376, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": 377, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": 378, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": 379, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": 380, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": 381, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": 382, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": 383, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": 384, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": 385, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": 386, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": 387, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": 388, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": 389, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": 390, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": 391, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": 392, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": 393, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": 394, + "p_model_model_language_model_embed_tokens_weight__rows_0_32768": 395, + "p_model_model_language_model_embed_tokens_weight__rows_32768_32768": 396, + "p_model_model_language_model_embed_tokens_weight__rows_65536_32768": 397, + "p_model_model_language_model_embed_tokens_weight__rows_98304_32768": 398, + "p_model_model_language_model_embed_tokens_weight__rows_131072_32768": 399, + "p_model_model_language_model_embed_tokens_weight__rows_163840_32768": 400, + "p_model_model_language_model_embed_tokens_weight__rows_196608_32768": 401, + "p_model_model_language_model_embed_tokens_weight__rows_229376_18944": 402, + "embedding__part_0": 403, + "embedding__part_1": 404, + "embedding__part_2": 405, + "embedding__part_3": 406, + "embedding__part_4": 407, + "embedding__part_5": 408, + "embedding__part_6": 409, + "embedding__part_7": 410, + "embedding__sum_0": 411, + "embedding__sum_1": 403, + "embedding__sum_2": 404, + "embedding__sum_3": 411, + "embedding__sum_4": 405, + "embedding__sum_5": 403, + "embedding": 406, + "unsqueeze": 404, + "expand": 407, + "slice_1": 411, + "unsqueeze_1": 408, + "unsqueeze_2": 405, + "unsqueeze_3": 409, + "expand_1": 403, + "unsqueeze_4": 410, + "_to_copy": 404, + "matmul": 407, + "transpose": 408, + "select": 405, + "select_1": 409, + "slice_2": 411, + "slice_3": 410, + "copy": 403, + "transpose_1": 404, + "select_2": 408, + "slice_scatter": 409, + "select_scatter": 405, + "transpose_2": 411, + "transpose_7": 407, + "select_7": 403, + "slice_7": 408, + "transpose_8": 409, + "select_8": 404, + "slice_8": 405, + "copy_1": 407, + "transpose_9": 403, + "select_9": 409, + "slice_scatter_1": 404, + "select_scatter_1": 405, + "transpose_10": 411, + "transpose_12": 407, + "select_11": 409, + "cat": 404, + "cos": 403, + "mul": 405, + "sin": 411, + "mul_1": 407, + "_to_copy_1": 409, + "_to_copy_2": 403, + "_to_copy_3": 412, + "pow_1": 413, + "mean": 404, + "add": 411, + "rsqrt": 405, + "mul_2": 413, + "_to_copy_4": 407, + "add_1": 404, + "mul_3": 412, + "type_as": 411, + "linear": 414, + "transpose_13": 415, + "linear_1": 405, + "view": 407, + "linear_2": 404, + "linear_3": 405, + "cat_1": 416, + "slice_10": 323, + "squeeze": 414, + "unsqueeze_5": 415, + "conv1d": 417, + "slice_11": 416, + "silu": 417, + "slice_12": 414, + "transpose_14": 415, + "slice_13": 411, + "slice_14": 413, + "slice_15": 412, + "view_1": 414, + "view_2": 411, + "view_3": 413, + "sigmoid": 412, + "_to_copy_5": 404, + "exp": 415, + "neg": 404, + "_to_copy_6": 415, + "add_2": 405, + "softplus": 415, + "mul_4": 405, + "mul_5": 404, + "sum_1": 415, + "add_3": 404, + "rsqrt_1": 415, + "mul_6": 404, + "mul_7": 414, + "sum_2": 415, + "add_4": 414, + "rsqrt_2": 415, + "mul_8": 414, + "_to_copy_7": 415, + "_to_copy_8": 416, + "_to_copy_9": 414, + "_to_copy_10": 411, + "normalized_gated_delta_rule": 418, + "slice_16": 324, + "slice_17": 413, + "transpose_15": 412, + "clone": 413, + "_to_copy_11": 404, + "view_4": 405, + "view_5": 411, + "_to_copy_12": 412, + "pow_2": 413, + "mean_1": 404, + "add_5": 407, + "rsqrt_3": 405, + "mul_9": 413, + "_to_copy_13": 404, + "mul_10": 407, + "_to_copy_14": 412, + "mul_11": 413, + "_to_copy_15": 405, + "view_6": 404, + "linear_4": 411, + "add_6": 407, + "_to_copy_16": 412, + "pow_3": 413, + "mean_2": 405, + "add_7": 404, + "rsqrt_4": 406, + "mul_12": 413, + "_to_copy_17": 411, + "add_8": 405, + "mul_13": 412, + "type_as_1": 404, + "linear_5": 415, + "linear_6": 414, + "mul_14": 416, + "linear_7": 406, + "add_9": 411, + "_to_copy_18": 413, + "pow_4": 412, + "mean_3": 405, + "add_10": 404, + "rsqrt_5": 407, + "mul_15": 412, + "_to_copy_19": 406, + "add_11": 405, + "mul_16": 413, + "type_as_2": 404, + "linear_8": 415, + "transpose_16": 414, + "linear_9": 407, + "view_7": 406, + "linear_10": 405, + "linear_11": 407, + "cat_2": 416, + "slice_18": 325, + "squeeze_1": 415, + "unsqueeze_6": 414, + "conv1d_1": 417, + "slice_19": 416, + "silu_3": 417, + "slice_20": 415, + "transpose_17": 414, + "slice_21": 404, + "slice_22": 412, + "slice_23": 413, + "view_8": 415, + "view_9": 404, + "view_10": 412, + "sigmoid_1": 413, + "_to_copy_20": 405, + "exp_1": 414, + "neg_1": 405, + "_to_copy_21": 414, + "add_12": 407, + "softplus_1": 414, + "mul_17": 407, + "mul_18": 405, + "sum_3": 414, + "add_13": 405, + "rsqrt_6": 414, + "mul_19": 405, + "mul_20": 415, + "sum_4": 414, + "add_14": 415, + "rsqrt_7": 414, + "mul_21": 415, + "_to_copy_22": 414, + "_to_copy_23": 416, + "_to_copy_24": 415, + "_to_copy_25": 404, + "normalized_gated_delta_rule_1": 418, + "slice_24": 326, + "slice_25": 412, + "transpose_18": 413, + "clone_1": 412, + "_to_copy_26": 405, + "view_11": 407, + "view_12": 404, + "_to_copy_27": 413, + "pow_5": 412, + "mean_4": 405, + "add_15": 406, + "rsqrt_8": 407, + "mul_22": 412, + "_to_copy_28": 405, + "mul_23": 406, + "_to_copy_29": 413, + "mul_24": 412, + "_to_copy_30": 407, + "view_13": 405, + "linear_12": 404, + "add_16": 406, + "_to_copy_31": 413, + "pow_6": 412, + "mean_5": 407, + "add_17": 405, + "rsqrt_9": 411, + "mul_25": 412, + "_to_copy_32": 404, + "add_18": 407, + "mul_26": 413, + "type_as_3": 405, + "linear_13": 414, + "linear_14": 415, + "mul_27": 416, + "linear_15": 411, + "add_19": 404, + "_to_copy_33": 412, + "pow_7": 413, + "mean_6": 407, + "add_20": 405, + "rsqrt_10": 406, + "mul_28": 413, + "_to_copy_34": 411, + "add_21": 407, + "mul_29": 412, + "type_as_4": 405, + "linear_16": 414, + "transpose_19": 415, + "linear_17": 406, + "view_14": 411, + "linear_18": 407, + "linear_19": 406, + "cat_3": 416, + "slice_26": 327, + "squeeze_2": 414, + "unsqueeze_7": 415, + "conv1d_2": 417, + "slice_27": 416, + "silu_6": 417, + "slice_28": 414, + "transpose_20": 415, + "slice_29": 405, + "slice_30": 413, + "slice_31": 412, + "view_15": 414, + "view_16": 405, + "view_17": 413, + "sigmoid_2": 412, + "_to_copy_35": 407, + "exp_2": 415, + "neg_2": 407, + "_to_copy_36": 415, + "add_22": 406, + "softplus_2": 415, + "mul_30": 406, + "mul_31": 407, + "sum_5": 415, + "add_23": 407, + "rsqrt_11": 415, + "mul_32": 407, + "mul_33": 414, + "sum_6": 415, + "add_24": 414, + "rsqrt_12": 415, + "mul_34": 414, + "_to_copy_37": 415, + "_to_copy_38": 416, + "_to_copy_39": 414, + "_to_copy_40": 405, + "normalized_gated_delta_rule_2": 418, + "slice_32": 328, + "slice_33": 413, + "transpose_21": 412, + "clone_2": 413, + "_to_copy_41": 407, + "view_18": 406, + "view_19": 405, + "_to_copy_42": 412, + "pow_8": 413, + "mean_7": 407, + "add_25": 411, + "rsqrt_13": 406, + "mul_35": 413, + "_to_copy_43": 407, + "mul_36": 411, + "_to_copy_44": 412, + "mul_37": 413, + "_to_copy_45": 406, + "view_20": 407, + "linear_20": 405, + "add_26": 411, + "_to_copy_46": 412, + "pow_9": 413, + "mean_8": 406, + "add_27": 407, + "rsqrt_14": 404, + "mul_38": 413, + "_to_copy_47": 405, + "add_28": 406, + "mul_39": 412, + "type_as_5": 407, + "linear_21": 415, + "linear_22": 414, + "mul_40": 416, + "linear_23": 404, + "add_29": 405, + "_to_copy_48": 413, + "pow_10": 412, + "mean_9": 406, + "add_30": 407, + "rsqrt_15": 411, + "mul_41": 412, + "_to_copy_49": 404, + "add_31": 406, + "mul_42": 413, + "type_as_6": 407, + "linear_24": 412, + "view_21": 413, + "slice_34": 411, + "slice_35": 404, + "clone_3": 406, + "_unsafe_view": 404, + "view_22": 406, + "_to_copy_50": 412, + "pow_11": 413, + "mean_10": 411, + "add_32": 406, + "rsqrt_16": 411, + "mul_43": 413, + "_to_copy_51": 406, + "add_33": 411, + "mul_44": 412, + "type_as_7": 406, + "transpose_22": 411, + "linear_25": 406, + "view_23": 413, + "_to_copy_52": 406, + "pow_12": 412, + "mean_11": 413, + "add_34": 412, + "rsqrt_17": 413, + "mul_45": 412, + "_to_copy_53": 406, + "add_35": 413, + "mul_46": 406, + "type_as_8": 412, + "transpose_23": 406, + "linear_26": 413, + "view_24": 407, + "transpose_24": 412, + "unsqueeze_8": 407, + "unsqueeze_9": 413, + "slice_36": 415, + "slice_37": 414, + "slice_38": 411, + "slice_39": 416, + "mul_47": 406, + "slice_40": 417, + "slice_41": 418, + "neg_3": 415, + "cat_4": 418, + "mul_48": 415, + "add_36": 417, + "mul_49": 406, + "slice_42": 407, + "slice_43": 415, + "neg_4": 411, + "cat_5": 415, + "mul_50": 411, + "add_37": 407, + "cat_6": 406, + "cat_7": 411, + "cache_update": 329, + "cache_update_1": 330, + "unsqueeze_10": 419, + "expand_2": 420, + "clone_4": 421, + "_unsafe_view_1": 420, + "unsqueeze_11": 419, + "expand_3": 421, + "clone_5": 422, + "_unsafe_view_2": 421, + "transpose_25": 422, + "matmul_1": 418, + "mul_51": 419, + "add_38": 418, + "softmax": 419, + "_to_copy_54": 418, + "matmul_2": 407, + "transpose_26": 411, + "clone_6": 406, + "view_25": 407, + "sigmoid_3": 411, + "mul_52": 406, + "linear_27": 404, + "add_39": 407, + "_to_copy_55": 413, + "pow_13": 412, + "mean_12": 411, + "add_40": 406, + "rsqrt_18": 405, + "mul_53": 412, + "_to_copy_56": 404, + "add_41": 411, + "mul_54": 413, + "type_as_9": 406, + "linear_28": 415, + "linear_29": 414, + "mul_55": 416, + "linear_30": 405, + "add_42": 404, + "_to_copy_57": 412, + "pow_14": 413, + "mean_13": 411, + "add_43": 406, + "rsqrt_19": 407, + "mul_56": 413, + "_to_copy_58": 405, + "add_44": 411, + "mul_57": 412, + "type_as_10": 406, + "linear_31": 415, + "transpose_27": 414, + "linear_32": 407, + "view_26": 405, + "linear_33": 411, + "linear_34": 407, + "cat_8": 416, + "slice_44": 331, + "squeeze_3": 415, + "unsqueeze_12": 414, + "conv1d_3": 417, + "slice_45": 416, + "silu_10": 417, + "slice_46": 415, + "transpose_28": 414, + "slice_47": 406, + "slice_48": 413, + "slice_49": 412, + "view_27": 415, + "view_28": 406, + "view_29": 413, + "sigmoid_4": 412, + "_to_copy_59": 411, + "exp_3": 414, + "neg_5": 411, + "_to_copy_60": 414, + "add_45": 407, + "softplus_3": 414, + "mul_58": 407, + "mul_59": 411, + "sum_7": 414, + "add_46": 411, + "rsqrt_20": 414, + "mul_60": 411, + "mul_61": 415, + "sum_8": 414, + "add_47": 415, + "rsqrt_21": 414, + "mul_62": 415, + "_to_copy_61": 414, + "_to_copy_62": 416, + "_to_copy_63": 415, + "_to_copy_64": 406, + "normalized_gated_delta_rule_3": 418, + "slice_50": 332, + "slice_51": 413, + "transpose_29": 412, + "clone_7": 413, + "_to_copy_65": 411, + "view_30": 407, + "view_31": 406, + "_to_copy_66": 412, + "pow_15": 413, + "mean_14": 411, + "add_48": 405, + "rsqrt_22": 407, + "mul_63": 413, + "_to_copy_67": 411, + "mul_64": 405, + "_to_copy_68": 412, + "mul_65": 413, + "_to_copy_69": 407, + "view_32": 411, + "linear_35": 406, + "add_49": 405, + "_to_copy_70": 412, + "pow_16": 413, + "mean_15": 407, + "add_50": 411, + "rsqrt_23": 404, + "mul_66": 413, + "_to_copy_71": 406, + "add_51": 407, + "mul_67": 412, + "type_as_11": 411, + "linear_36": 414, + "linear_37": 415, + "mul_68": 416, + "linear_38": 404, + "add_52": 406, + "_to_copy_72": 413, + "pow_17": 412, + "mean_16": 407, + "add_53": 411, + "rsqrt_24": 405, + "mul_69": 412, + "_to_copy_73": 404, + "add_54": 407, + "mul_70": 413, + "type_as_12": 411, + "linear_39": 414, + "transpose_30": 415, + "linear_40": 405, + "view_33": 404, + "linear_41": 407, + "linear_42": 405, + "cat_9": 416, + "slice_52": 333, + "squeeze_4": 414, + "unsqueeze_13": 415, + "conv1d_4": 417, + "slice_53": 416, + "silu_13": 417, + "slice_54": 414, + "transpose_31": 415, + "slice_55": 411, + "slice_56": 412, + "slice_57": 413, + "view_34": 414, + "view_35": 411, + "view_36": 412, + "sigmoid_5": 413, + "_to_copy_74": 407, + "exp_4": 415, + "neg_6": 407, + "_to_copy_75": 415, + "add_55": 405, + "softplus_4": 415, + "mul_71": 405, + "mul_72": 407, + "sum_9": 415, + "add_56": 407, + "rsqrt_25": 415, + "mul_73": 407, + "mul_74": 414, + "sum_10": 415, + "add_57": 414, + "rsqrt_26": 415, + "mul_75": 414, + "_to_copy_76": 415, + "_to_copy_77": 416, + "_to_copy_78": 414, + "_to_copy_79": 411, + "normalized_gated_delta_rule_4": 418, + "slice_58": 334, + "slice_59": 412, + "transpose_32": 413, + "clone_8": 412, + "_to_copy_80": 407, + "view_37": 405, + "view_38": 411, + "_to_copy_81": 413, + "pow_18": 412, + "mean_17": 407, + "add_58": 404, + "rsqrt_27": 405, + "mul_76": 412, + "_to_copy_82": 407, + "mul_77": 404, + "_to_copy_83": 413, + "mul_78": 412, + "_to_copy_84": 405, + "view_39": 407, + "linear_43": 411, + "add_59": 404, + "_to_copy_85": 413, + "pow_19": 412, + "mean_18": 405, + "add_60": 407, + "rsqrt_28": 406, + "mul_79": 412, + "_to_copy_86": 411, + "add_61": 405, + "mul_80": 413, + "type_as_13": 407, + "linear_44": 415, + "linear_45": 414, + "mul_81": 416, + "linear_46": 406, + "add_62": 411, + "_to_copy_87": 412, + "pow_20": 413, + "mean_19": 405, + "add_63": 407, + "rsqrt_29": 404, + "mul_82": 413, + "_to_copy_88": 406, + "add_64": 405, + "mul_83": 412, + "type_as_14": 407, + "linear_47": 415, + "transpose_33": 414, + "linear_48": 404, + "view_40": 406, + "linear_49": 405, + "linear_50": 404, + "cat_10": 416, + "slice_60": 335, + "squeeze_5": 415, + "unsqueeze_14": 414, + "conv1d_5": 417, + "slice_61": 416, + "silu_16": 417, + "slice_62": 415, + "transpose_34": 414, + "slice_63": 407, + "slice_64": 413, + "slice_65": 412, + "view_41": 415, + "view_42": 407, + "view_43": 413, + "sigmoid_6": 412, + "_to_copy_89": 405, + "exp_5": 414, + "neg_7": 405, + "_to_copy_90": 414, + "add_65": 404, + "softplus_5": 414, + "mul_84": 404, + "mul_85": 405, + "sum_11": 414, + "add_66": 405, + "rsqrt_30": 414, + "mul_86": 405, + "mul_87": 415, + "sum_12": 414, + "add_67": 415, + "rsqrt_31": 414, + "mul_88": 415, + "_to_copy_91": 414, + "_to_copy_92": 416, + "_to_copy_93": 415, + "_to_copy_94": 407, + "normalized_gated_delta_rule_5": 418, + "slice_66": 336, + "slice_67": 413, + "transpose_35": 412, + "clone_9": 413, + "_to_copy_95": 405, + "view_44": 404, + "view_45": 407, + "_to_copy_96": 412, + "pow_21": 413, + "mean_20": 405, + "add_68": 406, + "rsqrt_32": 404, + "mul_89": 413, + "_to_copy_97": 405, + "mul_90": 406, + "_to_copy_98": 412, + "mul_91": 413, + "_to_copy_99": 404, + "view_46": 405, + "linear_51": 407, + "add_69": 406, + "_to_copy_100": 412, + "pow_22": 413, + "mean_21": 404, + "add_70": 405, + "rsqrt_33": 411, + "mul_92": 413, + "_to_copy_101": 407, + "add_71": 404, + "mul_93": 412, + "type_as_15": 405, + "linear_52": 414, + "linear_53": 415, + "mul_94": 416, + "linear_54": 411, + "add_72": 407, + "_to_copy_102": 413, + "pow_23": 412, + "mean_22": 404, + "add_73": 405, + "rsqrt_34": 406, + "mul_95": 412, + "_to_copy_103": 411, + "add_74": 404, + "mul_96": 413, + "type_as_16": 405, + "linear_55": 412, + "view_47": 413, + "slice_68": 406, + "slice_69": 411, + "clone_10": 404, + "_unsafe_view_3": 411, + "view_48": 404, + "_to_copy_104": 412, + "pow_24": 413, + "mean_23": 406, + "add_75": 404, + "rsqrt_35": 406, + "mul_97": 413, + "_to_copy_105": 404, + "add_76": 406, + "mul_98": 412, + "type_as_17": 404, + "transpose_36": 406, + "linear_56": 404, + "view_49": 413, + "_to_copy_106": 404, + "pow_25": 412, + "mean_24": 413, + "add_77": 412, + "rsqrt_36": 413, + "mul_99": 412, + "_to_copy_107": 404, + "add_78": 413, + "mul_100": 404, + "type_as_18": 412, + "transpose_37": 404, + "linear_57": 413, + "view_50": 405, + "transpose_38": 412, + "unsqueeze_15": 405, + "unsqueeze_16": 413, + "slice_70": 414, + "slice_71": 415, + "slice_72": 406, + "slice_73": 416, + "mul_101": 404, + "slice_74": 417, + "slice_75": 418, + "neg_8": 414, + "cat_11": 418, + "mul_102": 414, + "add_79": 417, + "mul_103": 404, + "slice_76": 405, + "slice_77": 414, + "neg_9": 406, + "cat_12": 414, + "mul_104": 406, + "add_80": 405, + "cat_13": 404, + "cat_14": 406, + "cache_update_2": 337, + "cache_update_3": 338, + "unsqueeze_17": 419, + "expand_4": 420, + "clone_11": 422, + "_unsafe_view_4": 421, + "unsqueeze_18": 419, + "expand_5": 420, + "clone_12": 422, + "_unsafe_view_5": 420, + "transpose_39": 422, + "matmul_3": 418, + "mul_105": 419, + "add_81": 418, + "softmax_1": 419, + "_to_copy_108": 418, + "matmul_4": 405, + "transpose_40": 406, + "clone_13": 404, + "view_51": 405, + "sigmoid_7": 406, + "mul_106": 404, + "linear_58": 411, + "add_82": 405, + "_to_copy_109": 413, + "pow_26": 412, + "mean_25": 406, + "add_83": 404, + "rsqrt_37": 407, + "mul_107": 412, + "_to_copy_110": 411, + "add_84": 406, + "mul_108": 413, + "type_as_19": 404, + "linear_59": 414, + "linear_60": 415, + "mul_109": 416, + "linear_61": 407, + "add_85": 411, + "_to_copy_111": 412, + "pow_27": 413, + "mean_26": 406, + "add_86": 404, + "rsqrt_38": 405, + "mul_110": 413, + "_to_copy_112": 407, + "add_87": 406, + "mul_111": 412, + "type_as_20": 404, + "linear_62": 414, + "transpose_41": 415, + "linear_63": 405, + "view_52": 407, + "linear_64": 406, + "linear_65": 405, + "cat_15": 416, + "slice_78": 339, + "squeeze_6": 414, + "unsqueeze_19": 415, + "conv1d_6": 417, + "slice_79": 416, + "silu_20": 417, + "slice_80": 414, + "transpose_42": 415, + "slice_81": 404, + "slice_82": 413, + "slice_83": 412, + "view_53": 414, + "view_54": 404, + "view_55": 413, + "sigmoid_8": 412, + "_to_copy_113": 406, + "exp_6": 415, + "neg_10": 406, + "_to_copy_114": 415, + "add_88": 405, + "softplus_6": 415, + "mul_112": 405, + "mul_113": 406, + "sum_13": 415, + "add_89": 406, + "rsqrt_39": 415, + "mul_114": 406, + "mul_115": 414, + "sum_14": 415, + "add_90": 414, + "rsqrt_40": 415, + "mul_116": 414, + "_to_copy_115": 415, + "_to_copy_116": 416, + "_to_copy_117": 414, + "_to_copy_118": 404, + "normalized_gated_delta_rule_6": 418, + "slice_84": 340, + "slice_85": 413, + "transpose_43": 412, + "clone_14": 413, + "_to_copy_119": 406, + "view_56": 405, + "view_57": 404, + "_to_copy_120": 412, + "pow_28": 413, + "mean_27": 406, + "add_91": 407, + "rsqrt_41": 405, + "mul_117": 413, + "_to_copy_121": 406, + "mul_118": 407, + "_to_copy_122": 412, + "mul_119": 413, + "_to_copy_123": 405, + "view_58": 406, + "linear_66": 404, + "add_92": 407, + "_to_copy_124": 412, + "pow_29": 413, + "mean_28": 405, + "add_93": 406, + "rsqrt_42": 411, + "mul_120": 413, + "_to_copy_125": 404, + "add_94": 405, + "mul_121": 412, + "type_as_21": 406, + "linear_67": 415, + "linear_68": 414, + "mul_122": 416, + "linear_69": 411, + "add_95": 404, + "_to_copy_126": 413, + "pow_30": 412, + "mean_29": 405, + "add_96": 406, + "rsqrt_43": 407, + "mul_123": 412, + "_to_copy_127": 411, + "add_97": 405, + "mul_124": 413, + "type_as_22": 406, + "linear_70": 415, + "transpose_44": 414, + "linear_71": 407, + "view_59": 411, + "linear_72": 405, + "linear_73": 407, + "cat_16": 416, + "slice_86": 341, + "squeeze_7": 415, + "unsqueeze_20": 414, + "conv1d_7": 417, + "slice_87": 416, + "silu_23": 417, + "slice_88": 415, + "transpose_45": 414, + "slice_89": 406, + "slice_90": 412, + "slice_91": 413, + "view_60": 415, + "view_61": 406, + "view_62": 412, + "sigmoid_9": 413, + "_to_copy_128": 405, + "exp_7": 414, + "neg_11": 405, + "_to_copy_129": 414, + "add_98": 407, + "softplus_7": 414, + "mul_125": 407, + "mul_126": 405, + "sum_15": 414, + "add_99": 405, + "rsqrt_44": 414, + "mul_127": 405, + "mul_128": 415, + "sum_16": 414, + "add_100": 415, + "rsqrt_45": 414, + "mul_129": 415, + "_to_copy_130": 414, + "_to_copy_131": 416, + "_to_copy_132": 415, + "_to_copy_133": 406, + "normalized_gated_delta_rule_7": 418, + "slice_92": 342, + "slice_93": 412, + "transpose_46": 413, + "clone_15": 412, + "_to_copy_134": 405, + "view_63": 407, + "view_64": 406, + "_to_copy_135": 413, + "pow_31": 412, + "mean_30": 405, + "add_101": 411, + "rsqrt_46": 407, + "mul_130": 412, + "_to_copy_136": 405, + "mul_131": 411, + "_to_copy_137": 413, + "mul_132": 412, + "_to_copy_138": 407, + "view_65": 405, + "linear_74": 406, + "add_102": 411, + "_to_copy_139": 413, + "pow_32": 412, + "mean_31": 407, + "add_103": 405, + "rsqrt_47": 404, + "mul_133": 412, + "_to_copy_140": 406, + "add_104": 407, + "mul_134": 413, + "type_as_23": 405, + "linear_75": 414, + "linear_76": 415, + "mul_135": 416, + "linear_77": 404, + "add_105": 406, + "_to_copy_141": 412, + "pow_33": 413, + "mean_32": 407, + "add_106": 405, + "rsqrt_48": 411, + "mul_136": 413, + "_to_copy_142": 404, + "add_107": 407, + "mul_137": 412, + "type_as_24": 405, + "linear_78": 414, + "transpose_47": 415, + "linear_79": 411, + "view_66": 404, + "linear_80": 407, + "linear_81": 411, + "cat_17": 416, + "slice_94": 343, + "squeeze_8": 414, + "unsqueeze_21": 415, + "conv1d_8": 417, + "slice_95": 416, + "silu_26": 417, + "slice_96": 414, + "transpose_48": 415, + "slice_97": 405, + "slice_98": 413, + "slice_99": 412, + "view_67": 414, + "view_68": 405, + "view_69": 413, + "sigmoid_10": 412, + "_to_copy_143": 407, + "exp_8": 415, + "neg_12": 407, + "_to_copy_144": 415, + "add_108": 411, + "softplus_8": 415, + "mul_138": 411, + "mul_139": 407, + "sum_17": 415, + "add_109": 407, + "rsqrt_49": 415, + "mul_140": 407, + "mul_141": 414, + "sum_18": 415, + "add_110": 414, + "rsqrt_50": 415, + "mul_142": 414, + "_to_copy_145": 415, + "_to_copy_146": 416, + "_to_copy_147": 414, + "_to_copy_148": 405, + "normalized_gated_delta_rule_8": 418, + "slice_100": 344, + "slice_101": 413, + "transpose_49": 412, + "clone_16": 413, + "_to_copy_149": 407, + "view_70": 411, + "view_71": 405, + "_to_copy_150": 412, + "pow_34": 413, + "mean_33": 407, + "add_111": 404, + "rsqrt_51": 411, + "mul_143": 413, + "_to_copy_151": 407, + "mul_144": 404, + "_to_copy_152": 412, + "mul_145": 413, + "_to_copy_153": 411, + "view_72": 407, + "linear_82": 405, + "add_112": 404, + "_to_copy_154": 412, + "pow_35": 413, + "mean_34": 411, + "add_113": 407, + "rsqrt_52": 406, + "mul_146": 413, + "_to_copy_155": 405, + "add_114": 411, + "mul_147": 412, + "type_as_25": 407, + "linear_83": 415, + "linear_84": 414, + "mul_148": 416, + "linear_85": 406, + "add_115": 405, + "_to_copy_156": 413, + "pow_36": 412, + "mean_35": 411, + "add_116": 407, + "rsqrt_53": 404, + "mul_149": 412, + "_to_copy_157": 406, + "add_117": 411, + "mul_150": 413, + "type_as_26": 407, + "linear_86": 412, + "view_73": 413, + "slice_102": 404, + "slice_103": 406, + "clone_17": 411, + "_unsafe_view_6": 406, + "view_74": 411, + "_to_copy_158": 412, + "pow_37": 413, + "mean_36": 404, + "add_118": 411, + "rsqrt_54": 404, + "mul_151": 413, + "_to_copy_159": 411, + "add_119": 404, + "mul_152": 412, + "type_as_27": 411, + "transpose_50": 404, + "linear_87": 411, + "view_75": 413, + "_to_copy_160": 411, + "pow_38": 412, + "mean_37": 413, + "add_120": 412, + "rsqrt_55": 413, + "mul_153": 412, + "_to_copy_161": 411, + "add_121": 413, + "mul_154": 411, + "type_as_28": 412, + "transpose_51": 411, + "linear_88": 413, + "view_76": 407, + "transpose_52": 412, + "unsqueeze_22": 407, + "unsqueeze_23": 413, + "slice_104": 415, + "slice_105": 414, + "slice_106": 404, + "slice_107": 416, + "mul_155": 411, + "slice_108": 417, + "slice_109": 418, + "neg_13": 415, + "cat_18": 418, + "mul_156": 415, + "add_122": 417, + "mul_157": 411, + "slice_110": 407, + "slice_111": 415, + "neg_14": 404, + "cat_19": 415, + "mul_158": 404, + "add_123": 407, + "cat_20": 411, + "cat_21": 404, + "cache_update_4": 345, + "cache_update_5": 346, + "unsqueeze_24": 419, + "expand_6": 421, + "clone_18": 422, + "_unsafe_view_7": 420, + "unsqueeze_25": 419, + "expand_7": 421, + "clone_19": 422, + "_unsafe_view_8": 421, + "transpose_53": 422, + "matmul_5": 418, + "mul_159": 419, + "add_124": 418, + "softmax_2": 419, + "_to_copy_162": 418, + "matmul_6": 407, + "transpose_54": 404, + "clone_20": 411, + "view_77": 407, + "sigmoid_11": 404, + "mul_160": 411, + "linear_89": 406, + "add_125": 407, + "_to_copy_163": 413, + "pow_39": 412, + "mean_38": 404, + "add_126": 411, + "rsqrt_56": 405, + "mul_161": 412, + "_to_copy_164": 406, + "add_127": 404, + "mul_162": 413, + "type_as_29": 411, + "linear_90": 415, + "linear_91": 414, + "mul_163": 416, + "linear_92": 405, + "add_128": 406, + "_to_copy_165": 412, + "pow_40": 413, + "mean_39": 404, + "add_129": 411, + "rsqrt_57": 407, + "mul_164": 413, + "_to_copy_166": 405, + "add_130": 404, + "mul_165": 412, + "type_as_30": 411, + "linear_93": 415, + "transpose_55": 414, + "linear_94": 407, + "view_78": 405, + "linear_95": 404, + "linear_96": 407, + "cat_22": 416, + "slice_112": 347, + "squeeze_9": 415, + "unsqueeze_26": 414, + "conv1d_9": 417, + "slice_113": 416, + "silu_30": 417, + "slice_114": 415, + "transpose_56": 414, + "slice_115": 411, + "slice_116": 413, + "slice_117": 412, + "view_79": 415, + "view_80": 411, + "view_81": 413, + "sigmoid_12": 412, + "_to_copy_167": 404, + "exp_9": 414, + "neg_15": 404, + "_to_copy_168": 414, + "add_131": 407, + "softplus_9": 414, + "mul_166": 407, + "mul_167": 404, + "sum_19": 414, + "add_132": 404, + "rsqrt_58": 414, + "mul_168": 404, + "mul_169": 415, + "sum_20": 414, + "add_133": 415, + "rsqrt_59": 414, + "mul_170": 415, + "_to_copy_169": 414, + "_to_copy_170": 416, + "_to_copy_171": 415, + "_to_copy_172": 411, + "normalized_gated_delta_rule_9": 418, + "slice_118": 348, + "slice_119": 413, + "transpose_57": 412, + "clone_21": 413, + "_to_copy_173": 404, + "view_82": 407, + "view_83": 411, + "_to_copy_174": 412, + "pow_41": 413, + "mean_40": 404, + "add_134": 405, + "rsqrt_60": 407, + "mul_171": 413, + "_to_copy_175": 404, + "mul_172": 405, + "_to_copy_176": 412, + "mul_173": 413, + "_to_copy_177": 407, + "view_84": 404, + "linear_97": 411, + "add_135": 405, + "_to_copy_178": 412, + "pow_42": 413, + "mean_41": 407, + "add_136": 404, + "rsqrt_61": 406, + "mul_174": 413, + "_to_copy_179": 411, + "add_137": 407, + "mul_175": 412, + "type_as_31": 404, + "linear_98": 414, + "linear_99": 415, + "mul_176": 416, + "linear_100": 406, + "add_138": 411, + "_to_copy_180": 413, + "pow_43": 412, + "mean_42": 407, + "add_139": 404, + "rsqrt_62": 405, + "mul_177": 412, + "_to_copy_181": 406, + "add_140": 407, + "mul_178": 413, + "type_as_32": 404, + "linear_101": 414, + "transpose_58": 415, + "linear_102": 405, + "view_85": 406, + "linear_103": 407, + "linear_104": 405, + "cat_23": 416, + "slice_120": 349, + "squeeze_10": 414, + "unsqueeze_27": 415, + "conv1d_10": 417, + "slice_121": 416, + "silu_33": 417, + "slice_122": 414, + "transpose_59": 415, + "slice_123": 404, + "slice_124": 412, + "slice_125": 413, + "view_86": 414, + "view_87": 404, + "view_88": 412, + "sigmoid_13": 413, + "_to_copy_182": 407, + "exp_10": 415, + "neg_16": 407, + "_to_copy_183": 415, + "add_141": 405, + "softplus_10": 415, + "mul_179": 405, + "mul_180": 407, + "sum_21": 415, + "add_142": 407, + "rsqrt_63": 415, + "mul_181": 407, + "mul_182": 414, + "sum_22": 415, + "add_143": 414, + "rsqrt_64": 415, + "mul_183": 414, + "_to_copy_184": 415, + "_to_copy_185": 416, + "_to_copy_186": 414, + "_to_copy_187": 404, + "normalized_gated_delta_rule_10": 418, + "slice_126": 350, + "slice_127": 412, + "transpose_60": 413, + "clone_22": 412, + "_to_copy_188": 407, + "view_89": 405, + "view_90": 404, + "_to_copy_189": 413, + "pow_44": 412, + "mean_43": 407, + "add_144": 406, + "rsqrt_65": 405, + "mul_184": 412, + "_to_copy_190": 407, + "mul_185": 406, + "_to_copy_191": 413, + "mul_186": 412, + "_to_copy_192": 405, + "view_91": 407, + "linear_105": 404, + "add_145": 406, + "_to_copy_193": 413, + "pow_45": 412, + "mean_44": 405, + "add_146": 407, + "rsqrt_66": 411, + "mul_187": 412, + "_to_copy_194": 404, + "add_147": 405, + "mul_188": 413, + "type_as_33": 407, + "linear_106": 415, + "linear_107": 414, + "mul_189": 416, + "linear_108": 411, + "add_148": 404, + "_to_copy_195": 412, + "pow_46": 413, + "mean_45": 405, + "add_149": 407, + "rsqrt_67": 406, + "mul_190": 413, + "_to_copy_196": 411, + "add_150": 405, + "mul_191": 412, + "type_as_34": 407, + "linear_109": 415, + "transpose_61": 414, + "linear_110": 406, + "view_92": 411, + "linear_111": 405, + "linear_112": 406, + "cat_24": 416, + "slice_128": 351, + "squeeze_11": 415, + "unsqueeze_28": 414, + "conv1d_11": 417, + "slice_129": 416, + "silu_36": 417, + "slice_130": 415, + "transpose_62": 414, + "slice_131": 407, + "slice_132": 413, + "slice_133": 412, + "view_93": 415, + "view_94": 407, + "view_95": 413, + "sigmoid_14": 412, + "_to_copy_197": 405, + "exp_11": 414, + "neg_17": 405, + "_to_copy_198": 414, + "add_151": 406, + "softplus_11": 414, + "mul_192": 406, + "mul_193": 405, + "sum_23": 414, + "add_152": 405, + "rsqrt_68": 414, + "mul_194": 405, + "mul_195": 415, + "sum_24": 414, + "add_153": 415, + "rsqrt_69": 414, + "mul_196": 415, + "_to_copy_199": 414, + "_to_copy_200": 416, + "_to_copy_201": 415, + "_to_copy_202": 407, + "normalized_gated_delta_rule_11": 418, + "slice_134": 352, + "slice_135": 413, + "transpose_63": 412, + "clone_23": 413, + "_to_copy_203": 405, + "view_96": 406, + "view_97": 407, + "_to_copy_204": 412, + "pow_47": 413, + "mean_46": 405, + "add_154": 411, + "rsqrt_70": 406, + "mul_197": 413, + "_to_copy_205": 405, + "mul_198": 411, + "_to_copy_206": 412, + "mul_199": 413, + "_to_copy_207": 406, + "view_98": 405, + "linear_113": 407, + "add_155": 411, + "_to_copy_208": 412, + "pow_48": 413, + "mean_47": 406, + "add_156": 405, + "rsqrt_71": 404, + "mul_200": 413, + "_to_copy_209": 407, + "add_157": 406, + "mul_201": 412, + "type_as_35": 405, + "linear_114": 414, + "linear_115": 415, + "mul_202": 416, + "linear_116": 404, + "add_158": 407, + "_to_copy_210": 413, + "pow_49": 412, + "mean_48": 406, + "add_159": 405, + "rsqrt_72": 411, + "mul_203": 412, + "_to_copy_211": 404, + "add_160": 406, + "mul_204": 413, + "type_as_36": 405, + "linear_117": 412, + "view_99": 413, + "slice_136": 411, + "slice_137": 404, + "clone_24": 406, + "_unsafe_view_9": 404, + "view_100": 406, + "_to_copy_212": 412, + "pow_50": 413, + "mean_49": 411, + "add_161": 406, + "rsqrt_73": 411, + "mul_205": 413, + "_to_copy_213": 406, + "add_162": 411, + "mul_206": 412, + "type_as_37": 406, + "transpose_64": 411, + "linear_118": 406, + "view_101": 413, + "_to_copy_214": 406, + "pow_51": 412, + "mean_50": 413, + "add_163": 412, + "rsqrt_74": 413, + "mul_207": 412, + "_to_copy_215": 406, + "add_164": 413, + "mul_208": 406, + "type_as_38": 412, + "transpose_65": 406, + "linear_119": 413, + "view_102": 405, + "transpose_66": 412, + "unsqueeze_29": 405, + "unsqueeze_30": 413, + "slice_138": 414, + "slice_139": 415, + "slice_140": 411, + "slice_141": 416, + "mul_209": 406, + "slice_142": 417, + "slice_143": 418, + "neg_18": 414, + "cat_25": 418, + "mul_210": 414, + "add_165": 417, + "mul_211": 406, + "slice_144": 405, + "slice_145": 414, + "neg_19": 411, + "cat_26": 414, + "mul_212": 411, + "add_166": 405, + "cat_27": 406, + "cat_28": 411, + "cache_update_6": 353, + "cache_update_7": 354, + "unsqueeze_31": 419, + "expand_8": 420, + "clone_25": 422, + "_unsafe_view_10": 421, + "unsqueeze_32": 419, + "expand_9": 420, + "clone_26": 422, + "_unsafe_view_11": 420, + "transpose_67": 422, + "matmul_7": 418, + "mul_213": 419, + "add_167": 418, + "softmax_3": 419, + "_to_copy_216": 418, + "matmul_8": 405, + "transpose_68": 411, + "clone_27": 406, + "view_103": 405, + "sigmoid_15": 411, + "mul_214": 406, + "linear_120": 404, + "add_168": 405, + "_to_copy_217": 413, + "pow_52": 412, + "mean_51": 411, + "add_169": 406, + "rsqrt_75": 407, + "mul_215": 412, + "_to_copy_218": 404, + "add_170": 411, + "mul_216": 413, + "type_as_39": 406, + "linear_121": 414, + "linear_122": 415, + "mul_217": 416, + "linear_123": 407, + "add_171": 404, + "_to_copy_219": 412, + "pow_53": 413, + "mean_52": 411, + "add_172": 406, + "rsqrt_76": 405, + "mul_218": 413, + "_to_copy_220": 407, + "add_173": 411, + "mul_219": 412, + "type_as_40": 406, + "linear_124": 414, + "transpose_69": 415, + "linear_125": 405, + "view_104": 407, + "linear_126": 411, + "linear_127": 405, + "cat_29": 416, + "slice_146": 355, + "squeeze_12": 414, + "unsqueeze_33": 415, + "conv1d_12": 417, + "slice_147": 416, + "silu_40": 417, + "slice_148": 414, + "transpose_70": 415, + "slice_149": 406, + "slice_150": 413, + "slice_151": 412, + "view_105": 414, + "view_106": 406, + "view_107": 413, + "sigmoid_16": 412, + "_to_copy_221": 411, + "exp_12": 415, + "neg_20": 411, + "_to_copy_222": 415, + "add_174": 405, + "softplus_12": 415, + "mul_220": 405, + "mul_221": 411, + "sum_25": 415, + "add_175": 411, + "rsqrt_77": 415, + "mul_222": 411, + "mul_223": 414, + "sum_26": 415, + "add_176": 414, + "rsqrt_78": 415, + "mul_224": 414, + "_to_copy_223": 415, + "_to_copy_224": 416, + "_to_copy_225": 414, + "_to_copy_226": 406, + "normalized_gated_delta_rule_12": 418, + "slice_152": 356, + "slice_153": 413, + "transpose_71": 412, + "clone_28": 413, + "_to_copy_227": 411, + "view_108": 405, + "view_109": 406, + "_to_copy_228": 412, + "pow_54": 413, + "mean_53": 411, + "add_177": 407, + "rsqrt_79": 405, + "mul_225": 413, + "_to_copy_229": 411, + "mul_226": 407, + "_to_copy_230": 412, + "mul_227": 413, + "_to_copy_231": 405, + "view_110": 411, + "linear_128": 406, + "add_178": 407, + "_to_copy_232": 412, + "pow_55": 413, + "mean_54": 405, + "add_179": 411, + "rsqrt_80": 404, + "mul_228": 413, + "_to_copy_233": 406, + "add_180": 405, + "mul_229": 412, + "type_as_41": 411, + "linear_129": 415, + "linear_130": 414, + "mul_230": 416, + "linear_131": 404, + "add_181": 406, + "_to_copy_234": 413, + "pow_56": 412, + "mean_55": 405, + "add_182": 411, + "rsqrt_81": 407, + "mul_231": 412, + "_to_copy_235": 404, + "add_183": 405, + "mul_232": 413, + "type_as_42": 411, + "linear_132": 415, + "transpose_72": 414, + "linear_133": 407, + "view_111": 404, + "linear_134": 405, + "linear_135": 407, + "cat_30": 416, + "slice_154": 357, + "squeeze_13": 415, + "unsqueeze_34": 414, + "conv1d_13": 417, + "slice_155": 416, + "silu_43": 417, + "slice_156": 415, + "transpose_73": 414, + "slice_157": 411, + "slice_158": 412, + "slice_159": 413, + "view_112": 415, + "view_113": 411, + "view_114": 412, + "sigmoid_17": 413, + "_to_copy_236": 405, + "exp_13": 414, + "neg_21": 405, + "_to_copy_237": 414, + "add_184": 407, + "softplus_13": 414, + "mul_233": 407, + "mul_234": 405, + "sum_27": 414, + "add_185": 405, + "rsqrt_82": 414, + "mul_235": 405, + "mul_236": 415, + "sum_28": 414, + "add_186": 415, + "rsqrt_83": 414, + "mul_237": 415, + "_to_copy_238": 414, + "_to_copy_239": 416, + "_to_copy_240": 415, + "_to_copy_241": 411, + "normalized_gated_delta_rule_13": 418, + "slice_160": 358, + "slice_161": 412, + "transpose_74": 413, + "clone_29": 412, + "_to_copy_242": 405, + "view_115": 407, + "view_116": 411, + "_to_copy_243": 413, + "pow_57": 412, + "mean_56": 405, + "add_187": 404, + "rsqrt_84": 407, + "mul_238": 412, + "_to_copy_244": 405, + "mul_239": 404, + "_to_copy_245": 413, + "mul_240": 412, + "_to_copy_246": 407, + "view_117": 405, + "linear_136": 411, + "add_188": 404, + "_to_copy_247": 413, + "pow_58": 412, + "mean_57": 407, + "add_189": 405, + "rsqrt_85": 406, + "mul_241": 412, + "_to_copy_248": 411, + "add_190": 407, + "mul_242": 413, + "type_as_43": 405, + "linear_137": 414, + "linear_138": 415, + "mul_243": 416, + "linear_139": 406, + "add_191": 411, + "_to_copy_249": 412, + "pow_59": 413, + "mean_58": 407, + "add_192": 405, + "rsqrt_86": 404, + "mul_244": 413, + "_to_copy_250": 406, + "add_193": 407, + "mul_245": 412, + "type_as_44": 405, + "linear_140": 414, + "transpose_75": 415, + "linear_141": 404, + "view_118": 406, + "linear_142": 407, + "linear_143": 404, + "cat_31": 416, + "slice_162": 359, + "squeeze_14": 414, + "unsqueeze_35": 415, + "conv1d_14": 417, + "slice_163": 416, + "silu_46": 417, + "slice_164": 414, + "transpose_76": 415, + "slice_165": 405, + "slice_166": 413, + "slice_167": 412, + "view_119": 414, + "view_120": 405, + "view_121": 413, + "sigmoid_18": 412, + "_to_copy_251": 407, + "exp_14": 415, + "neg_22": 407, + "_to_copy_252": 415, + "add_194": 404, + "softplus_14": 415, + "mul_246": 404, + "mul_247": 407, + "sum_29": 415, + "add_195": 407, + "rsqrt_87": 415, + "mul_248": 407, + "mul_249": 414, + "sum_30": 415, + "add_196": 414, + "rsqrt_88": 415, + "mul_250": 414, + "_to_copy_253": 415, + "_to_copy_254": 416, + "_to_copy_255": 414, + "_to_copy_256": 405, + "normalized_gated_delta_rule_14": 418, + "slice_168": 360, + "slice_169": 413, + "transpose_77": 412, + "clone_30": 413, + "_to_copy_257": 407, + "view_122": 404, + "view_123": 405, + "_to_copy_258": 412, + "pow_60": 413, + "mean_59": 407, + "add_197": 406, + "rsqrt_89": 404, + "mul_251": 413, + "_to_copy_259": 407, + "mul_252": 406, + "_to_copy_260": 412, + "mul_253": 413, + "_to_copy_261": 404, + "view_124": 407, + "linear_144": 405, + "add_198": 406, + "_to_copy_262": 412, + "pow_61": 413, + "mean_60": 404, + "add_199": 407, + "rsqrt_90": 411, + "mul_254": 413, + "_to_copy_263": 405, + "add_200": 404, + "mul_255": 412, + "type_as_45": 407, + "linear_145": 415, + "linear_146": 414, + "mul_256": 416, + "linear_147": 411, + "add_201": 405, + "_to_copy_264": 413, + "pow_62": 412, + "mean_61": 404, + "add_202": 407, + "rsqrt_91": 406, + "mul_257": 412, + "_to_copy_265": 411, + "add_203": 404, + "mul_258": 413, + "type_as_46": 407, + "linear_148": 412, + "view_125": 413, + "slice_170": 406, + "slice_171": 411, + "clone_31": 404, + "_unsafe_view_12": 411, + "view_126": 404, + "_to_copy_266": 412, + "pow_63": 413, + "mean_62": 406, + "add_204": 404, + "rsqrt_92": 406, + "mul_259": 413, + "_to_copy_267": 404, + "add_205": 406, + "mul_260": 412, + "type_as_47": 404, + "transpose_78": 406, + "linear_149": 404, + "view_127": 413, + "_to_copy_268": 404, + "pow_64": 412, + "mean_63": 413, + "add_206": 412, + "rsqrt_93": 413, + "mul_261": 412, + "_to_copy_269": 404, + "add_207": 413, + "mul_262": 404, + "type_as_48": 412, + "transpose_79": 404, + "linear_150": 413, + "view_128": 407, + "transpose_80": 412, + "unsqueeze_36": 407, + "unsqueeze_37": 413, + "slice_172": 415, + "slice_173": 414, + "slice_174": 406, + "slice_175": 416, + "mul_263": 404, + "slice_176": 417, + "slice_177": 418, + "neg_23": 415, + "cat_32": 418, + "mul_264": 415, + "add_208": 417, + "mul_265": 404, + "slice_178": 407, + "slice_179": 415, + "neg_24": 406, + "cat_33": 415, + "mul_266": 406, + "add_209": 407, + "cat_34": 404, + "cat_35": 406, + "cache_update_8": 361, + "cache_update_9": 362, + "unsqueeze_38": 419, + "expand_10": 421, + "clone_32": 422, + "_unsafe_view_13": 420, + "unsqueeze_39": 419, + "expand_11": 421, + "clone_33": 422, + "_unsafe_view_14": 421, + "transpose_81": 422, + "matmul_9": 418, + "mul_267": 419, + "add_210": 418, + "softmax_4": 419, + "_to_copy_270": 418, + "matmul_10": 407, + "transpose_82": 406, + "clone_34": 404, + "view_129": 407, + "sigmoid_19": 406, + "mul_268": 404, + "linear_151": 411, + "add_211": 407, + "_to_copy_271": 413, + "pow_65": 412, + "mean_64": 406, + "add_212": 404, + "rsqrt_94": 405, + "mul_269": 412, + "_to_copy_272": 411, + "add_213": 406, + "mul_270": 413, + "type_as_49": 404, + "linear_152": 415, + "linear_153": 414, + "mul_271": 416, + "linear_154": 405, + "add_214": 411, + "_to_copy_273": 412, + "pow_66": 413, + "mean_65": 406, + "add_215": 404, + "rsqrt_95": 407, + "mul_272": 413, + "_to_copy_274": 405, + "add_216": 406, + "mul_273": 412, + "type_as_50": 404, + "linear_155": 415, + "transpose_83": 414, + "linear_156": 407, + "view_130": 405, + "linear_157": 406, + "linear_158": 407, + "cat_36": 416, + "slice_180": 363, + "squeeze_15": 415, + "unsqueeze_40": 414, + "conv1d_15": 417, + "slice_181": 416, + "silu_50": 417, + "slice_182": 415, + "transpose_84": 414, + "slice_183": 404, + "slice_184": 413, + "slice_185": 412, + "view_131": 415, + "view_132": 404, + "view_133": 413, + "sigmoid_20": 412, + "_to_copy_275": 406, + "exp_15": 414, + "neg_25": 406, + "_to_copy_276": 414, + "add_217": 407, + "softplus_15": 414, + "mul_274": 407, + "mul_275": 406, + "sum_31": 414, + "add_218": 406, + "rsqrt_96": 414, + "mul_276": 406, + "mul_277": 415, + "sum_32": 414, + "add_219": 415, + "rsqrt_97": 414, + "mul_278": 415, + "_to_copy_277": 414, + "_to_copy_278": 416, + "_to_copy_279": 415, + "_to_copy_280": 404, + "normalized_gated_delta_rule_15": 418, + "slice_186": 364, + "slice_187": 413, + "transpose_85": 412, + "clone_35": 413, + "_to_copy_281": 406, + "view_134": 407, + "view_135": 404, + "_to_copy_282": 412, + "pow_67": 413, + "mean_66": 406, + "add_220": 405, + "rsqrt_98": 407, + "mul_279": 413, + "_to_copy_283": 406, + "mul_280": 405, + "_to_copy_284": 412, + "mul_281": 413, + "_to_copy_285": 407, + "view_136": 406, + "linear_159": 404, + "add_221": 405, + "_to_copy_286": 412, + "pow_68": 413, + "mean_67": 407, + "add_222": 406, + "rsqrt_99": 411, + "mul_282": 413, + "_to_copy_287": 404, + "add_223": 407, + "mul_283": 412, + "type_as_51": 406, + "linear_160": 414, + "linear_161": 415, + "mul_284": 416, + "linear_162": 411, + "add_224": 404, + "_to_copy_288": 413, + "pow_69": 412, + "mean_68": 407, + "add_225": 406, + "rsqrt_100": 405, + "mul_285": 412, + "_to_copy_289": 411, + "add_226": 407, + "mul_286": 413, + "type_as_52": 406, + "linear_163": 414, + "transpose_86": 415, + "linear_164": 405, + "view_137": 411, + "linear_165": 407, + "linear_166": 405, + "cat_37": 416, + "slice_188": 365, + "squeeze_16": 414, + "unsqueeze_41": 415, + "conv1d_16": 417, + "slice_189": 416, + "silu_53": 417, + "slice_190": 414, + "transpose_87": 415, + "slice_191": 406, + "slice_192": 412, + "slice_193": 413, + "view_138": 414, + "view_139": 406, + "view_140": 412, + "sigmoid_21": 413, + "_to_copy_290": 407, + "exp_16": 415, + "neg_26": 407, + "_to_copy_291": 415, + "add_227": 405, + "softplus_16": 415, + "mul_287": 405, + "mul_288": 407, + "sum_33": 415, + "add_228": 407, + "rsqrt_101": 415, + "mul_289": 407, + "mul_290": 414, + "sum_34": 415, + "add_229": 414, + "rsqrt_102": 415, + "mul_291": 414, + "_to_copy_292": 415, + "_to_copy_293": 416, + "_to_copy_294": 414, + "_to_copy_295": 406, + "normalized_gated_delta_rule_16": 418, + "slice_194": 366, + "slice_195": 412, + "transpose_88": 413, + "clone_36": 412, + "_to_copy_296": 407, + "view_141": 405, + "view_142": 406, + "_to_copy_297": 413, + "pow_70": 412, + "mean_69": 407, + "add_230": 411, + "rsqrt_103": 405, + "mul_292": 412, + "_to_copy_298": 407, + "mul_293": 411, + "_to_copy_299": 413, + "mul_294": 412, + "_to_copy_300": 405, + "view_143": 407, + "linear_167": 406, + "add_231": 411, + "_to_copy_301": 413, + "pow_71": 412, + "mean_70": 405, + "add_232": 407, + "rsqrt_104": 404, + "mul_295": 412, + "_to_copy_302": 406, + "add_233": 405, + "mul_296": 413, + "type_as_53": 407, + "linear_168": 415, + "linear_169": 414, + "mul_297": 416, + "linear_170": 404, + "add_234": 406, + "_to_copy_303": 412, + "pow_72": 413, + "mean_71": 405, + "add_235": 407, + "rsqrt_105": 411, + "mul_298": 413, + "_to_copy_304": 404, + "add_236": 405, + "mul_299": 412, + "type_as_54": 407, + "linear_171": 415, + "transpose_89": 414, + "linear_172": 411, + "view_144": 404, + "linear_173": 405, + "linear_174": 411, + "cat_38": 416, + "slice_196": 367, + "squeeze_17": 415, + "unsqueeze_42": 414, + "conv1d_17": 417, + "slice_197": 416, + "silu_56": 417, + "slice_198": 415, + "transpose_90": 414, + "slice_199": 407, + "slice_200": 413, + "slice_201": 412, + "view_145": 415, + "view_146": 407, + "view_147": 413, + "sigmoid_22": 412, + "_to_copy_305": 405, + "exp_17": 414, + "neg_27": 405, + "_to_copy_306": 414, + "add_237": 411, + "softplus_17": 414, + "mul_300": 411, + "mul_301": 405, + "sum_35": 414, + "add_238": 405, + "rsqrt_106": 414, + "mul_302": 405, + "mul_303": 415, + "sum_36": 414, + "add_239": 415, + "rsqrt_107": 414, + "mul_304": 415, + "_to_copy_307": 414, + "_to_copy_308": 416, + "_to_copy_309": 415, + "_to_copy_310": 407, + "normalized_gated_delta_rule_17": 418, + "slice_202": 368, + "slice_203": 413, + "transpose_91": 412, + "clone_37": 413, + "_to_copy_311": 405, + "view_148": 411, + "view_149": 407, + "_to_copy_312": 412, + "pow_73": 413, + "mean_72": 405, + "add_240": 404, + "rsqrt_108": 411, + "mul_305": 413, + "_to_copy_313": 405, + "mul_306": 404, + "_to_copy_314": 412, + "mul_307": 413, + "_to_copy_315": 411, + "view_150": 405, + "linear_175": 407, + "add_241": 404, + "_to_copy_316": 412, + "pow_74": 413, + "mean_73": 411, + "add_242": 405, + "rsqrt_109": 406, + "mul_308": 413, + "_to_copy_317": 407, + "add_243": 411, + "mul_309": 412, + "type_as_55": 405, + "linear_176": 414, + "linear_177": 415, + "mul_310": 416, + "linear_178": 406, + "add_244": 407, + "_to_copy_318": 413, + "pow_75": 412, + "mean_74": 411, + "add_245": 405, + "rsqrt_110": 404, + "mul_311": 412, + "_to_copy_319": 406, + "add_246": 411, + "mul_312": 413, + "type_as_56": 405, + "linear_179": 412, + "view_151": 413, + "slice_204": 404, + "slice_205": 406, + "clone_38": 411, + "_unsafe_view_15": 406, + "view_152": 411, + "_to_copy_320": 412, + "pow_76": 413, + "mean_75": 404, + "add_247": 411, + "rsqrt_111": 404, + "mul_313": 413, + "_to_copy_321": 411, + "add_248": 404, + "mul_314": 412, + "type_as_57": 411, + "transpose_92": 404, + "linear_180": 411, + "view_153": 413, + "_to_copy_322": 411, + "pow_77": 412, + "mean_76": 413, + "add_249": 412, + "rsqrt_112": 413, + "mul_315": 412, + "_to_copy_323": 411, + "add_250": 413, + "mul_316": 411, + "type_as_58": 412, + "transpose_93": 411, + "linear_181": 413, + "view_154": 405, + "transpose_94": 412, + "unsqueeze_43": 405, + "unsqueeze_44": 409, + "slice_206": 403, + "slice_207": 413, + "slice_208": 404, + "slice_209": 414, + "mul_317": 411, + "slice_210": 415, + "slice_211": 416, + "neg_28": 403, + "cat_39": 416, + "mul_318": 403, + "add_251": 415, + "mul_319": 411, + "slice_212": 403, + "slice_213": 405, + "neg_29": 404, + "cat_40": 405, + "mul_320": 404, + "add_252": 403, + "cat_41": 405, + "cat_42": 409, + "cache_update_10": 369, + "cache_update_11": 370, + "unsqueeze_45": 419, + "expand_12": 420, + "clone_39": 422, + "_unsafe_view_16": 421, + "unsqueeze_46": 419, + "expand_13": 420, + "clone_40": 422, + "_unsafe_view_17": 420, + "transpose_95": 422, + "matmul_11": 418, + "mul_321": 419, + "add_253": 418, + "softmax_5": 419, + "_to_copy_324": 418, + "matmul_12": 411, + "transpose_96": 404, + "clone_41": 403, + "view_155": 409, + "sigmoid_23": 405, + "mul_322": 411, + "linear_182": 404, + "add_254": 403, + "_to_copy_325": 413, + "pow_78": 412, + "mean_77": 406, + "add_255": 409, + "rsqrt_113": 405, + "mul_323": 412, + "_to_copy_326": 411, + "add_256": 407, + "mul_324": 413, + "type_as_59": 404, + "linear_183": 415, + "linear_184": 414, + "mul_325": 416, + "linear_185": 406, + "add_257": 409, + "_to_copy_327": 412, + "pow_79": 413, + "mean_78": 405, + "add_258": 411, + "rsqrt_114": 407, + "mul_326": 413, + "_to_copy_328": 404, + "add_259": 403, + "mul_327": 412, + "type_as_60": 406, + "alias": 409, + "linear_186__part_0": 418, + "linear_186__part_1": 419, + "linear_186__part_2": 421, + "linear_186__part_3": 422, + "linear_186__part_4": 420, + "linear_186__part_5": 423, + "linear_186__part_6": 424, + "linear_186__part_7": 425, + "linear_186__cat_0_0": 426, + "linear_186__cat_0_7": 423, + "linear_186": 419 + }, + "byteLength": 3590446880 + }, + "rewrites": [ + { + "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": "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": 32768, + "bindingLimitBytes": 134217728 + }, + { + "node": "linear_186", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 32768, + "bindingLimitBytes": 134217728 + } + ] +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/graph.json b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/graph.json new file mode 100644 index 0000000000000000000000000000000000000000..57645002748cf5ac5fd16fb93846d1249815fe33 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/graph.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee2e4f0cd06a4ac73213766c544fb6cfca8e2cf39104961bc7469ea3c5c0c6a2 +size 11991700 diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6255820b113009706aa9f477eac21208f3b230a5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/002f3dff1b027a15b1b92e3f04c88efa1ecd7199d6852603412b700547bdc70d.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..57f490e587dab593560daa3c9772c2b9140361b7 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/023d7b2d9e71a92f74db0e4583d453118915f46ba0c8c96c796632d958654ea1.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 >= 8u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/02915a0f22b3245bc6490c9bfbedc331f50293e3e7f87525fe4482f0c94e8aff.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/02915a0f22b3245bc6490c9bfbedc331f50293e3e7f87525fe4482f0c94e8aff.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..219b8a3d094d939e6b6f6c1c99b7a05e0cd0ebf7 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/02915a0f22b3245bc6490c9bfbedc331f50293e3e7f87525fe4482f0c94e8aff.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 1536u) % 1u) * 2048u + ((i / 192u) % 8u) * 256u + ((i / 192u) % 1u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/02b89cf7780686b8cf46c4db784144cef988e047230cf92ecd1e25e45cabd224.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/02b89cf7780686b8cf46c4db784144cef988e047230cf92ecd1e25e45cabd224.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c4b3fc8c9f4326ec8fc257af3939c29e4440bd3b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/02b89cf7780686b8cf46c4db784144cef988e047230cf92ecd1e25e45cabd224.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(f32(b0[i]) * 0.0625); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0c3e9d8766b1f4555d042d2b65a4523aec844512 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/035979e9647fb0752d2135ad95461bae23933ee3fc591c65963104942287d12f.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 2048u + col; + var acc = 0.0; + for (var tile = 0u; tile < 6144u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 6144u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 6144u + tile + local.x]); } + if (group.x * 8u + local.y < 2048u && tile + local.x < 6144u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 6144u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 2048u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f2f4362e1c4570f8072a48668537582cd1cb5e24 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/037256742da99812bfb6619450f2af103bf17e3d38333c9d309f07aba3a3ae20.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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; } +let x = f32(b0[i]); +out[i] = f16(x / (1.0 + exp(-x))); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/03d3272d16bc278fcc977e95fcb512bac183f503c7d6ced229f9f01f8e5011bf.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/03d3272d16bc278fcc977e95fcb512bac183f503c7d6ced229f9f01f8e5011bf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..17c6b71bc1f39f0dd464a4663ccaf87152bac894 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/03d3272d16bc278fcc977e95fcb512bac183f503c7d6ced229f9f01f8e5011bf.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 0 || token >= 32768) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 0) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/044fcdf70bcfbb2e3fa472f622959964b1e5929797fcd65fa2eab7016910ee34.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/044fcdf70bcfbb2e3fa472f622959964b1e5929797fcd65fa2eab7016910ee34.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..786b63d2cb6a3c8e0107e6a9a10242e4a1e32c10 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/044fcdf70bcfbb2e3fa472f622959964b1e5929797fcd65fa2eab7016910ee34.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 18944u; + if (i >= 18944u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 18944u) * 2048u + p]) * f32(b1[(i % 18944u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/04f26661cc9ac73bb7f39f66b76c4be31ee8774e094eeb41ed7d677e1b1d3fa9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/04f26661cc9ac73bb7f39f66b76c4be31ee8774e094eeb41ed7d677e1b1d3fa9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..67133bd5fb075efe996a921acde0d02717ffa6f3 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e7d9d234b05b94c10577eaeabcc515dfcf64cb5c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/05a8da37281cc1281b5928ab86a013be4a4648b3a422652a3f525829a8981036.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 >= 32u) { return; } +var acc = 0.0; +for (var j = 0u; j < 256u; j++) { acc += f32(b0[(((i / 32u) % 1u) * 8192u + ((i / 8u) % 4u) * 2048u + ((i / 1u) % 8u) * 256u) + (((j / 1u) % 256u) * 1u)]); } +out[i] = f32(acc / 256.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0725908abf78c022f86deb338a06779f6f1346ebfe54ff5a43ee6b9eeb93c972.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0725908abf78c022f86deb338a06779f6f1346ebfe54ff5a43ee6b9eeb93c972.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..71b292e22afb334a345d95567bc425387a266f55 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/078c19606137dad9d459e8041919bac37c31e51702a6d466d4d3df095a6d1f5a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/078c19606137dad9d459e8041919bac37c31e51702a6d466d4d3df095a6d1f5a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6c29f26a28b9e919e850e6744ec2cf9800e1c846 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/078c19606137dad9d459e8041919bac37c31e51702a6d466d4d3df095a6d1f5a.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 32768u; + if (i >= 32768u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 32768u) * 2048u + p]) * f32(b1[(i % 32768u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fce7e987ece305a3bb4becb659c06fc1ad5e5960 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0818b2d5166084f4c42b7d514426619af003f77e54291df8f45b72f91dfe5736.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 256u) % 1u) * 512u + ((i / 128u) % 2u) * 256u + ((i / 32u) % 4u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0835fa80eaadb7c0e6d677e1fc622ad5dcedabb431e5d312f3b7aeb2252a7c7c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0835fa80eaadb7c0e6d677e1fc622ad5dcedabb431e5d312f3b7aeb2252a7c7c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..91f740aa7006d14dca451533719861c5823ff016 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..13d326cdb9a9319b0b0441dab09bce0e7ba5080e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/08476f486255074c433aee58f82b8ec4150571dc774680f5d8b183a0a7e63e18.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 8192u) % 1u) * 24576u + ((i / 2048u) % 4u) * 6144u + (((i / 1u) % 2048u) * 1u + 2048u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0888235d9d0cb75cbdb598cbb2d235a2db071bed0dd8193164f8f21eab5a8ba9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0888235d9d0cb75cbdb598cbb2d235a2db071bed0dd8193164f8f21eab5a8ba9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cf8c932790b6de3696da8abd6a49e7fa0a9a4f24 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0888235d9d0cb75cbdb598cbb2d235a2db071bed0dd8193164f8f21eab5a8ba9.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) + (f32(b1[((i / 1u) % 4096u) * 1u]) * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/09e236dd54b3abd945c2d4fef393e9e9810f0d90cdec0d5deafd68ed3b95083d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/09e236dd54b3abd945c2d4fef393e9e9810f0d90cdec0d5deafd68ed3b95083d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..efefa9875fba543b9ab32306f58ba55c9f3e31dd --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/09e236dd54b3abd945c2d4fef393e9e9810f0d90cdec0d5deafd68ed3b95083d.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 / 256u) % 2u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/09e48926914d3759efd11d83aa3ba539f06f33c6f2c28fcc0c12811f9cd1bc44.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/09e48926914d3759efd11d83aa3ba539f06f33c6f2c28fcc0c12811f9cd1bc44.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c81c6fe894f9ccfb287f2822ffc25de0ed3b6d89 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/09e48926914d3759efd11d83aa3ba539f06f33c6f2c28fcc0c12811f9cd1bc44.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..957e22d690670bcf12444e81c17b680e37dffc82 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0a9068c3d7c002ffe2631245f0873bf56d726b5d933b040851e20cd8fad9fe1a.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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 * 4194240u; +if (i >= 8388608u) { return; } +out[i] = f16(b0[((i / 4194304u) % 2u) * 1048576u + ((i / 256u) % 4096u) * 256u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0b8815b1efe176d55ab08530c0a9e1913cf6c9ef3e773c574843573d237cf196.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0b8815b1efe176d55ab08530c0a9e1913cf6c9ef3e773c574843573d237cf196.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..686ab91f45e66184b49d0f2e7429a2171e634bad --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0d3892883d1532d173a7dbe85d9ab1daa87f0d9b4923a535b8d2003f24f9500c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0d3892883d1532d173a7dbe85d9ab1daa87f0d9b4923a535b8d2003f24f9500c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8520366c6db35027de02453b6c726e0fc678ab4f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0da9c6227714e2a6c031928cb6a6db229c1cb0a53a9309aea4e833a9b2b1f2f1.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0da9c6227714e2a6c031928cb6a6db229c1cb0a53a9309aea4e833a9b2b1f2f1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..85ec37d559d666fdce0ebfacea55cf92fb1309a5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0da9c6227714e2a6c031928cb6a6db229c1cb0a53a9309aea4e833a9b2b1f2f1.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 64u) % 1u) * 128u + ((i / 32u) % 2u) * 64u + ((i / 32u) % 1u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0fbed2c2803544176a2948c7666e91d918788652 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0e6c14d2118fe801932c1b41995a65c8e165bf8df3989e40793f509448988218.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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) % 256u; +if (coord >= 0u && coord < 64u) { out[i] = f16(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f16(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0ecd1d59b46e658df21ec09d4256b75bf9a57138f7bfe5e2bcc309a352a37468.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0ecd1d59b46e658df21ec09d4256b75bf9a57138f7bfe5e2bcc309a352a37468.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e7352812d12df3db1f28020dd00baac41ac032f3 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0ecd1d59b46e658df21ec09d4256b75bf9a57138f7bfe5e2bcc309a352a37468.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 131072 || token >= 163840) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 131072) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..15c550b1f32e4f6bb9a861ae384749b50592bd5d --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0fb8222de35c84fb88807d4a3880f21188cf48d64c89a21aa208d7d72a48c273.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0fb8222de35c84fb88807d4a3880f21188cf48d64c89a21aa208d7d72a48c273.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1950645a0a8045e09651a52cb7794360cc5a81e1 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0fb8222de35c84fb88807d4a3880f21188cf48d64c89a21aa208d7d72a48c273.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0ff3d9cacb67aa460769fcf57dcfe10c5354e7e98681a4647a9acf1045d60892.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0ff3d9cacb67aa460769fcf57dcfe10c5354e7e98681a4647a9acf1045d60892.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cce3c1f1db7b7f4e9ed1c96b2333b4f17fe3421f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/0ff3d9cacb67aa460769fcf57dcfe10c5354e7e98681a4647a9acf1045d60892.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 2048u; + if (i >= 2048u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 6144u; p += 64u) { acc += f32(b0[(i / 2048u) * 6144u + p]) * f32(b1[(i % 2048u) * 6144u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/11a92e9cba6bfc381c98c5127db137639d6c87f1cb220459880ead039dc81256.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/11a92e9cba6bfc381c98c5127db137639d6c87f1cb220459880ead039dc81256.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..34af207bc7ce034721ed448af19bbd095c4bdb49 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4977dd26d91c116b0df884dba3354bba5ff74661 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/13dc73935b9b3b978232043e13babee3877bd53ca572be440de0590dacce8b17.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/13dc73935b9b3b978232043e13babee3877bd53ca572be440de0590dacce8b17.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..de89efb15b0a288331488e51bd02760c3dded87f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1545025e0a625ccc73b42ae112c6f9d78a1504c10b759979ae2247984f25f056.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1545025e0a625ccc73b42ae112c6f9d78a1504c10b759979ae2247984f25f056.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..94be53f6bd2811f15de48160517e54acd7fde791 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1608c6cf00b5863ad002f089be58fce76cff2c61fd43e3bb54ba90add6f41b02.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1608c6cf00b5863ad002f089be58fce76cff2c61fd43e3bb54ba90add6f41b02.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c87c9e1944c99c07b07b84e15fea0c9fc0f2334a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1608c6cf00b5863ad002f089be58fce76cff2c61fd43e3bb54ba90add6f41b02.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(-x); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/16757bdfacc94a92cfc58c9a3ba83ca377b8ea7ddf82fa8f6611f797a6934f14.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/16757bdfacc94a92cfc58c9a3ba83ca377b8ea7ddf82fa8f6611f797a6934f14.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d993098942267cb79d876b3e1dda448aa4b7720f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/16757bdfacc94a92cfc58c9a3ba83ca377b8ea7ddf82fa8f6611f797a6934f14.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 32768 || token >= 65536) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 32768) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4858e20d8cd328d8402a94f32bb7b90b01196533 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/16de135b328f53749dc017ed0a624ea63f39df86947f72a0d5d4aa06bedeadd1.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/17d9b9ec53d19b9ce8e3e750fd7beeb668dbc872a19e0ab5d29e566d84490e74.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/17d9b9ec53d19b9ce8e3e750fd7beeb668dbc872a19e0ab5d29e566d84490e74.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..917b26cf6f0184a1f5be451b35464be56dd4cbc0 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/17d9b9ec53d19b9ce8e3e750fd7beeb668dbc872a19e0ab5d29e566d84490e74.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 2048u; + if (i >= 2048u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 2048u) * 2048u + p]) * f32(b1[(i % 2048u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/17ec6bdaf98edf7657b74a25036c2bec806ecbb4280f33fff6afd2e0de5ed377.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/17ec6bdaf98edf7657b74a25036c2bec806ecbb4280f33fff6afd2e0de5ed377.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f8b87c0bd2b9575bbbc8395e5d566f057aeb69dd --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..460942834da66dc168f11860dfbd38c9cd069806 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..41fd5a021bacdb59aa4b8e651a7ca5fc0a5156ce --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1a3351b3225f169402dd4c0b1bfbcc407f8b5fdccfb10c74ccb80b7c3bd85a37.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1a3351b3225f169402dd4c0b1bfbcc407f8b5fdccfb10c74ccb80b7c3bd85a37.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d9439b84d9f7da36e78721572cf7161a2fd355e9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1a3351b3225f169402dd4c0b1bfbcc407f8b5fdccfb10c74ccb80b7c3bd85a37.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 196608 || token >= 229376) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 196608) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5c5819710be5b6147bca047695faeb86145b5c06 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1c7201caab74d50198d8bc7168ba7da0c97ece74a5ebdd791f88d6cfb5086975.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 >= 8u) { return; } +out[i] = f32(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8a0172f9f2a14fef750ae0b6360b798f12261529 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1cb76debd822c58d652ea0e82d2f5bdebc3dd67a8cf540154fd85f504fc72cd2.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 8192u) % 1u) * 8192u + ((i / 2048u) % 4u) * 256u + ((i / 256u) % 8u) * 1024u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a9dc55baa3fe224cff4c4abecf49782efc6b1d0f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1ce543c34cc4033e22afd4fc00a62ed19f44307cc4aeb52e7a503538e5831e9d.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d7d92b0e10d57438b8c7ca361df42da3284ffb4a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1df96eddb194a002a44f5a610245f84d192a38d2fff876c30fd55a4f518e8b25.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(-x); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f92885c31e275850573f7b7466c6f5db5311a385 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1f252e46a7c6dc933462fc88d57b832b6e87ffe79232ad52b4dad28e5083cc69.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(f32(b0[i]) * 0.0625); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..82fd5c53373876475fc6d1c7c107d991af93714a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/1f821653e02930a221ffe24970bbbd13b88c5e7bd119af97f624b075b1ab6534.wgsl @@ -0,0 +1,14 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/222a2645e3e02d0a4063a63ca1b6fa4c89faee6b76905b20de4dd54ab3077ea3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/222a2645e3e02d0a4063a63ca1b6fa4c89faee6b76905b20de4dd54ab3077ea3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1447aa7dfa8f3765ba1ecc1e492c6fce42213779 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/234987ec68497956ea7aef8e3f44c8a73dbe4c1519165d123946bedbb4d18e26.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/234987ec68497956ea7aef8e3f44c8a73dbe4c1519165d123946bedbb4d18e26.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f6c451610050d5332f2e96fe27fb0fc3d8359841 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/234987ec68497956ea7aef8e3f44c8a73dbe4c1519165d123946bedbb4d18e26.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 / 128u) % 16u) * 128u + 0u * 128u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/26655d5856c1c506477f9af344dc55197aa9220fe6e447eb38b11add83e278f3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/26655d5856c1c506477f9af344dc55197aa9220fe6e447eb38b11add83e278f3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4fe9e773a38d7021e6a43f1fb09471620ef28cd2 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/26655d5856c1c506477f9af344dc55197aa9220fe6e447eb38b11add83e278f3.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/26862ccab76186d81a45f06c5a7c9e1cd04e4592dbe151341043725a49a5cd2f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/26862ccab76186d81a45f06c5a7c9e1cd04e4592dbe151341043725a49a5cd2f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..777855e9a0619a31265d40e40b978418274c1301 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/26862ccab76186d81a45f06c5a7c9e1cd04e4592dbe151341043725a49a5cd2f.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 24576u) % 1u) * 30720u + ((i / 4u) % 6144u) * 5u + (((i / 1u) % 4u) * 1u + 1u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2711a4a396071f4e99303d01d4854ccaa37a445f4b718ebd7b61cf1c37e56b48.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2711a4a396071f4e99303d01d4854ccaa37a445f4b718ebd7b61cf1c37e56b48.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8a950952abaa36f3937eab9390ad3807c223cb72 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2711a4a396071f4e99303d01d4854ccaa37a445f4b718ebd7b61cf1c37e56b48.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] = f32(b0[((i / 16u) % 1u) * 16u + ((i / 1u) % 16u) * 1u + ((i / 1u) % 1u) * 16u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..807c82713fd489c5c7715967264c5919e02e7ce8 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/280b909ef0d849a03152d7314ee264a5c56fc7bf059fc775f1c288d9e59d2f45.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 / 1u) % 2048u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..09eab9f70d6fa16fc6be5f4092a3ce42c0d4ccdb --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/28fefba4b037ddf39d186e7eb1bfb1d5326df5ed875017cedcae7dbb2b502a0c.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/291830b3cb692bab94344428e329c3dc9417a43f48e1c42e414cf3f956686f1f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/291830b3cb692bab94344428e329c3dc9417a43f48e1c42e414cf3f956686f1f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a9f0acecf8ccd82d99e363498b3e79e3017fd782 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/291830b3cb692bab94344428e329c3dc9417a43f48e1c42e414cf3f956686f1f.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[((i / 1u) % 128u) * 1u]) * f32(b1[i])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b874e9a4efed02fbf5df29eda03d2b9a15ebba9c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2b707696a2df65d49bff973579e864e32bec86f16b559093c983e36ccae31a4a.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2c0cca2a7b0198c317e68992e64f3fe9bba80b67204c2695c20c372471c150d3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2c0cca2a7b0198c317e68992e64f3fe9bba80b67204c2695c20c372471c150d3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..05c73e3e7e256b462ade35867855db48ad811b19 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2c6b48ecf46e2d5956447e131ed77da3b567090ba8f22e1942b8b88366b84cee.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2c6b48ecf46e2d5956447e131ed77da3b567090ba8f22e1942b8b88366b84cee.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8966e8a80f3f44bdd01b47700f67d03f9249f211 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2c6b48ecf46e2d5956447e131ed77da3b567090ba8f22e1942b8b88366b84cee.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 98304 || token >= 131072) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 98304) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..38310e6def55923b47e15db35ea758fb1a108a62 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2d720c8b479e2b9ad67a833f8a8fa994f07aa2ab391c5dae833e368c9ef9c630.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2d720c8b479e2b9ad67a833f8a8fa994f07aa2ab391c5dae833e368c9ef9c630.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fc2cef636a402044d111ced0fa30ea7fec4456d9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2d720c8b479e2b9ad67a833f8a8fa994f07aa2ab391c5dae833e368c9ef9c630.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 < 229376u) { out[i] = f16(b0[(i / 248320u) * 229376u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 229376u && coord < 248320u) { out[i] = f16(b1[(i / 248320u) * 18944u + (coord - 229376u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..89a11494cd7f33adda4e076ebdbfd70d81c7c326 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2db8a995aaf1e89bf2da258d8e77c4ccc53a4d61ab9b352757da0e5a7321729d.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 * 2048u; +if (i >= 2048u) { return; } +let x = f32(b0[i]); +out[i] = f32(x*x); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2dd29b61b039fcdd7fbc3c94f629b24b6f81aa5b9923098837597359b295e40d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2dd29b61b039fcdd7fbc3c94f629b24b6f81aa5b9923098837597359b295e40d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9548ed074fc5f4321188aa402aa399a4525a86b3 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2dd29b61b039fcdd7fbc3c94f629b24b6f81aa5b9923098837597359b295e40d.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 / 2048u) % 1u) * 128u + ((i / 128u) % 16u) * 128u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2e1ce6ef01198780289fd4a35d8f86b5172595ed189cf931e8444091d49e8d3a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2e1ce6ef01198780289fd4a35d8f86b5172595ed189cf931e8444091d49e8d3a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2269a620c58e84a4903bb1135733ef2385d6af46 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2ecb3b88f329d2d66b34465ebc62482b3a9d40d37c69a30d55b69d2c85f082cb.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2ecb3b88f329d2d66b34465ebc62482b3a9d40d37c69a30d55b69d2c85f082cb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3c63ce2f0dca0db14d9ea99d0b7961a8e3ca077d --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/2ecb3b88f329d2d66b34465ebc62482b3a9d40d37c69a30d55b69d2c85f082cb.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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; } +let x = f32(b0[i]); +out[i] = f16(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d26f55dc801072951721a041127533286d4ae6fe --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3203e08f2c99fa4848f553739b7115a7c677840344f17b1d4391d6130517f5a7.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3203e08f2c99fa4848f553739b7115a7c677840344f17b1d4391d6130517f5a7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c50d890cb53585ca0b01fc04aaf5f843d4872ab1 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/32a5027cac5dd969d63fe5d5c88d7c87abb600c43ad1d70e1c56615ea781eff3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/32a5027cac5dd969d63fe5d5c88d7c87abb600c43ad1d70e1c56615ea781eff3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..acbc0b91943f7202976d3a0119d1ad2f45215e18 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/344153d8e20e545c4656ce08b7236cd65878f5ea91014447f4a129ea94025970.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/344153d8e20e545c4656ce08b7236cd65878f5ea91014447f4a129ea94025970.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..72c41ba6267d60e0c0315a6fec0eb455ae06c942 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/344153d8e20e545c4656ce08b7236cd65878f5ea91014447f4a129ea94025970.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 256u) % 1u) * 512u + ((i / 32u) % 8u) * 64u + ((i / 32u) % 1u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..067602c863a4ab63f4b880cb7b0d9a8b2fd5f4f5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/36bdc2a39d74582b6dc45860ff236f186c9f1643c68f6e6db2c48be5f8177d99.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[((i / 1u) % 128u) * 1u]) * f32(b1[i])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6e14292187b8316a8c107ad2f62fdffd925daec6 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/370bab74beead2c5e9cc8893585f79ae97ee5eace554cee12555a8ac8293b1c0.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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 * 4194240u; +if (i >= 8388608u) { return; } +out[i] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/377880fa04b2ba2c6238045a88750ff214243eb739603a04b16b69221ec5a6c2.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/377880fa04b2ba2c6238045a88750ff214243eb739603a04b16b69221ec5a6c2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..211fd6598a97833eb4f50fd80a39c9b7f4f50be7 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/377880fa04b2ba2c6238045a88750ff214243eb739603a04b16b69221ec5a6c2.wgsl @@ -0,0 +1,18 @@ +enable f16; +@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 batch = i / 4096u; +let row = (i / 4096u) % 1u; +let col = i % 4096u; +var acc = 0.0; +for (var p = 0u; p < 256u; p++) { + acc += f32(b0[(((batch / 1u) % 8u) * 256u) + row * 256u + p]) * f32(b1[(((batch / 1u) % 8u) * 1048576u) + p * 4096u + col]); +} +out[i] = f16(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/378330e4bcfd0c2f194f2ca261988c352ba4cdffd71dee16af52c3b7ba25657f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/378330e4bcfd0c2f194f2ca261988c352ba4cdffd71dee16af52c3b7ba25657f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..327312e60caf98bc40d9b0ce6fd14c6f02848d13 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/378330e4bcfd0c2f194f2ca261988c352ba4cdffd71dee16af52c3b7ba25657f.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 2048u + ((i / 2048u) % 1u) * 256u + ((i / 256u) % 8u) * 256u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c036f05ea84471b133dff547fd59dd9346fac3bd --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3976659e4383cb6f1c4f9c58ccef28215f246d506dae2455e18f51a78217549d.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 >= 4u) { return; } +var acc = 0.0; +for (var j = 0u; j < 2048u; j++) { acc += f32(b0[(((i / 4u) % 1u) * 8192u + ((i / 1u) % 4u) * 2048u) + (((j / 1u) % 2048u) * 1u)]); } +out[i] = f32(acc / 2048.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..03b0ac03a5a914b6b68e84517d7978c5b0914e51 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3a525a2cf0ad4be5cb90d14cfedbbd6814a3e42d8b1e66015f524ed9cfa407e8.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 24576u) % 1u) * 24576u + ((i / 6144u) % 4u) * 1u + ((i / 1u) % 6144u) * 4u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3a87037580f20ca38623bdcc88abd77f8cc3a83d8b3d1726756a1ec01656a496.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3a87037580f20ca38623bdcc88abd77f8cc3a83d8b3d1726756a1ec01656a496.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ad74b3648b0ea1f8c1b65017eabec261d2e554fa --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3a87037580f20ca38623bdcc88abd77f8cc3a83d8b3d1726756a1ec01656a496.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 6144u) % 1u) * 6144u + ((i / 1u) % 6144u) * 1u + ((i / 1u) % 1u) * 6144u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3bb850bda95bf77706a2839616a401b773f9de0428d1b427e61fb04e2c35fe74.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3bb850bda95bf77706a2839616a401b773f9de0428d1b427e61fb04e2c35fe74.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a3e680fcc3fb413659faf5613327f7d216420af9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3f78504435af26dbd6b62c6720a640528ce172b2e9062bd4d09806e3d2553423.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3f78504435af26dbd6b62c6720a640528ce172b2e9062bd4d09806e3d2553423.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..52fe147802d5ce29b2c5bc1180fa9e1abd1f2baf --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3f78504435af26dbd6b62c6720a640528ce172b2e9062bd4d09806e3d2553423.wgsl @@ -0,0 +1,14 @@ +enable f16; +@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 token = (i / 256u) % 1u; + let outer = i / 256u; + let destination = outer * 1048576u + u32(b1[token]) * 256u + i % 256u; + out[((destination) / 256u) % 4096u * 512u + ((destination) / 1048576u) % 2u * 256u + ((destination) / 1u) % 256u * 1u] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..76fef58bda8c3635f8de713e9d2e8807bca9fe33 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3fa1efae6d90015ba69cb40cef009f49ad7f7a960018fb92a962707deff4f09e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3fa1efae6d90015ba69cb40cef009f49ad7f7a960018fb92a962707deff4f09e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dd0f6856d5316f3d610290ed98cf1a9c465f554d --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3fe8cec0f9e8b563387f62876982a8f87cb5f03a1030cedde9cdb8e88c52e100.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3fe8cec0f9e8b563387f62876982a8f87cb5f03a1030cedde9cdb8e88c52e100.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5b91258b195c002afb3f851f977a360497cbf63e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3fe8cec0f9e8b563387f62876982a8f87cb5f03a1030cedde9cdb8e88c52e100.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 4096u; + if (i >= 4096u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 4096u) * 2048u + p]) * f32(b1[(i % 4096u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3ff863808627e030edefbb05f743b5a2cc19412b22779508a9d173a40d2ecc6e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3ff863808627e030edefbb05f743b5a2cc19412b22779508a9d173a40d2ecc6e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c075d6dfe88061f343f951725253abb44c30c4fb --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/3ff863808627e030edefbb05f743b5a2cc19412b22779508a9d173a40d2ecc6e.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 >= 2u) { return; } +var acc = 0.0; +for (var j = 0u; j < 256u; j++) { acc += f32(b0[(((i / 2u) % 1u) * 512u + ((i / 2u) % 1u) * 512u + ((i / 1u) % 2u) * 256u) + (((j / 1u) % 256u) * 1u)]); } +out[i] = f32(acc / 256.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/408b4edf1721b552fd1406f682e9c1612b2662372860bf14e040f011c1ee0cb9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/408b4edf1721b552fd1406f682e9c1612b2662372860bf14e040f011c1ee0cb9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..00b61866bc91528d842be55ceab8361398a8ffcc --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/408b4edf1721b552fd1406f682e9c1612b2662372860bf14e040f011c1ee0cb9.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b1[((i / 128u) % 16u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ad33c85df483204905cbf44f632d017ab01d900e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/40a960da8ca3bfbbbd5d4de7adf52321e356cff55ba70936dcabe1cb005a1b3f.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c49ee62d7457a49841eb532dbe4b892e1206d616 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/416cfd9be1617183bbf005e964b8bb5259b87f2cf184935f2dbb934afc3ca037.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 6144u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 6144u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 6144u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/421e5b9b6315c1f131e8e555ee8d1d23751643d1d071e48fa177bcc2f65d3712.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/421e5b9b6315c1f131e8e555ee8d1d23751643d1d071e48fa177bcc2f65d3712.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9fb463b8a07721f317043244653bea4782e4a01a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/421e5b9b6315c1f131e8e555ee8d1d23751643d1d071e48fa177bcc2f65d3712.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 / 1u) % 2048u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4282aa48d394c7c4b0d6c6f1063526bda4b9da2708f77a1945793a4614107efb.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4282aa48d394c7c4b0d6c6f1063526bda4b9da2708f77a1945793a4614107efb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..34e454679b05cf533c851743edc32e5b5a62a5ca --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4282aa48d394c7c4b0d6c6f1063526bda4b9da2708f77a1945793a4614107efb.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 4096u + ((i / 2048u) % 1u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 256u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/42c5af4838daf75fff9d4bb9a5df409511c1d6daed617d05eaf033a977503eac.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/42c5af4838daf75fff9d4bb9a5df409511c1d6daed617d05eaf033a977503eac.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ab4b18691ec8371839272e82e1191ffc00772c73 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/44a50c95f6df65dd90f18b3425789a3b213a25a0e3c80080df677f2972aa63e9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/44a50c95f6df65dd90f18b3425789a3b213a25a0e3c80080df677f2972aa63e9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..91e50f6753615da374819f9562f40758e4f1ad8b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/44a50c95f6df65dd90f18b3425789a3b213a25a0e3c80080df677f2972aa63e9.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 0 || token >= 32768) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 0) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4540d87eaec90c9eb42badfe0abfa70252f26b6e7de4dd31ee6903fe50e96bd1.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4540d87eaec90c9eb42badfe0abfa70252f26b6e7de4dd31ee6903fe50e96bd1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a0caa5c08df286ab5dbba084e1b6b90da802badc --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/45bb598b61c05eb6cdf699755e2877c598bb6fda0eee019e60d8301535dd3ec2.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/45bb598b61c05eb6cdf699755e2877c598bb6fda0eee019e60d8301535dd3ec2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..7e46591c2e309d554498e54fd9b83191549fe136 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/45bb598b61c05eb6cdf699755e2877c598bb6fda0eee019e60d8301535dd3ec2.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/48f34a333ff70495892101c428694fcddedec29f954b659093d206abbe1f7c15.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/48f34a333ff70495892101c428694fcddedec29f954b659093d206abbe1f7c15.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..98e80452fa7114465d478458819d5c9d107883d5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/48f34a333ff70495892101c428694fcddedec29f954b659093d206abbe1f7c15.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 >= 1u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dc3d6070aa02c4dad5f2bf7125c9300ad9860ca8 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4a952811352a76ed9369f2e11ba4d52917b618a5bc95e3232fdfa0579a79bef3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4a952811352a76ed9369f2e11ba4d52917b618a5bc95e3232fdfa0579a79bef3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5616870154be4b18b79eaafbb181a936311e6b40 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4a952811352a76ed9369f2e11ba4d52917b618a5bc95e3232fdfa0579a79bef3.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b0[i])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4be963f9a4bb928764fd0efe02769e220b0e69defb2f8b22ab28e41a90d3fc14.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4be963f9a4bb928764fd0efe02769e220b0e69defb2f8b22ab28e41a90d3fc14.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..24b926b3b907f9052d46718955950614e3f4d5b2 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4cd438e7ca1c026269bc19fdf9fb0fdf0041d106ab97f168b3ee259159689b08.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4cd438e7ca1c026269bc19fdf9fb0fdf0041d106ab97f168b3ee259159689b08.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9848d49d87bf035f2bf9eec381b221828165ff58 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4cd438e7ca1c026269bc19fdf9fb0fdf0041d106ab97f168b3ee259159689b08.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[0u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4e730849e680dcd1773c63d6bdb3398b869163b7aca5ca81ae12445b805ba7af.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4e730849e680dcd1773c63d6bdb3398b869163b7aca5ca81ae12445b805ba7af.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4d2049fc9c50b4face344269ce87d844b539411a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/4e730849e680dcd1773c63d6bdb3398b869163b7aca5ca81ae12445b805ba7af.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 163840 || token >= 196608) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 163840) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/506e74c3370bbc59adb7e554bbb95deaee0213f0dd19a4942ecb9a36f5dde1a4.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/506e74c3370bbc59adb7e554bbb95deaee0213f0dd19a4942ecb9a36f5dde1a4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..196699f7df19709c404310ff82383e9ae60bc748 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/506e74c3370bbc59adb7e554bbb95deaee0213f0dd19a4942ecb9a36f5dde1a4.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 98304 || token >= 131072) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 98304) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/509a7d601af6ce0b40a93d88c99825ea5d7ba01a258bd52ddad3eb10d2bb2804.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/509a7d601af6ce0b40a93d88c99825ea5d7ba01a258bd52ddad3eb10d2bb2804.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e1ff3093bd8eadd6d322b52148f556149e5e9b80 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/509a7d601af6ce0b40a93d88c99825ea5d7ba01a258bd52ddad3eb10d2bb2804.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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 * 18944u; +if (i >= 18944u) { return; } +let coord = (i / 1u) % 18944u; +if (coord >= 0u && coord < 18944u) { out[i] = f16(b0[(i / 18944u) * 18944u + (coord - 0u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..95189a197a347cc835d16d08e03cba68fa4c2bd3 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/50a2aaffa2f6e9866a2456ff60aa2a54cb817b737210774ef3ec1ba64834954e.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 / 1u) % 256u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fc751a3c5de6830de0104027f986f119ec7276f2 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/52b439507af68e969a6e116d3c8ea5338a4cde5a3fc1e97b729512d17ab2f4f9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/52b439507af68e969a6e116d3c8ea5338a4cde5a3fc1e97b729512d17ab2f4f9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..67ce4656ea03be5b73e9bffd1ef4a470f3754605 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/53d6eb0b5e6d49bfd3bb9fd1b781e13317adf528e047da2416e95f5a3a13b1b4.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/53d6eb0b5e6d49bfd3bb9fd1b781e13317adf528e047da2416e95f5a3a13b1b4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b42167010260a49ed2060dcc92a7440b82333b7e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/53d6eb0b5e6d49bfd3bb9fd1b781e13317adf528e047da2416e95f5a3a13b1b4.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 >= 96u) { return; } +out[i] = f32(b0[((i / 32u) % 3u) * 32u + ((i / 32u) % 1u) * 32u + ((i / 32u) % 1u) * 1u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5471ae38e894829494bbf27a5b01ac3256a61849dc2e23cdc190fd63954b8cd8.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5471ae38e894829494bbf27a5b01ac3256a61849dc2e23cdc190fd63954b8cd8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0aa75ecbf1242e3ff93afe626715479ed01d650d --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5471ae38e894829494bbf27a5b01ac3256a61849dc2e23cdc190fd63954b8cd8.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 >= 2u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fed6cc7462668b3acfb8722e9d56c54bfbc85de6 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5685f28b7fe39e87763365d0b4e311eb35a07f4728106448a936b739a785d878.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 >= 8u) { return; } +var acc = 0.0; +for (var j = 0u; j < 256u; j++) { acc += f32(b0[(((i / 8u) % 1u) * 2048u + ((i / 2u) % 4u) * 512u + ((i / 1u) % 2u) * 256u) + (((j / 1u) % 256u) * 1u)]); } +out[i] = f32(acc / 256.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/570f05a5f9f892e7a8075cde8b7bee28bc921f6054167eec4b85797f15b873da.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/570f05a5f9f892e7a8075cde8b7bee28bc921f6054167eec4b85797f15b873da.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c322e5ceb165920d18214506bc0a9b7cd148a40c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/570f05a5f9f892e7a8075cde8b7bee28bc921f6054167eec4b85797f15b873da.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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 * 75776u; +if (i >= 75776u) { return; } +let coord = (i / 1u) % 18944u; +if (coord >= 0u && coord < 18944u) { out[i] = f16(b0[(i / 18944u) * 18944u + (coord - 0u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e837c12551206937555f7bc9b7c86ebccdd92353 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/578ba670650363664652f61f499af0f982bea5e68d2448225f4311b76bc6754b.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 32768u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 32768u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 32768u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/578c3f0d7eb85a27b4da58e79e8842b993125d088e11e214d9d20be447388d80.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/578c3f0d7eb85a27b4da58e79e8842b993125d088e11e214d9d20be447388d80.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d6728c1ee46190528ad1c6e3036ab76b97b971a5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/578c3f0d7eb85a27b4da58e79e8842b993125d088e11e214d9d20be447388d80.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; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 1u) % 16u) * 128u) + (((j / 1u) % 128u) * 1u)]); } +out[i] = f32(acc / 128.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/57dc406322bf3badd8490493ef924bd7e7e0afd542d2d6e409b4d8a3b4b784f7.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/57dc406322bf3badd8490493ef924bd7e7e0afd542d2d6e409b4d8a3b4b784f7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5442fb5e440bbbbc406fb5988a08504f70fb6527 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/57dc406322bf3badd8490493ef924bd7e7e0afd542d2d6e409b4d8a3b4b784f7.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] = f32(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/58ef8a5a09de45f92e4696b0ce6e26de2500f935d75049d78111d723fdce3780.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/58ef8a5a09de45f92e4696b0ce6e26de2500f935d75049d78111d723fdce3780.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c61f2bf7efea5153ce0a52d836939ee53984b862 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/58ef8a5a09de45f92e4696b0ce6e26de2500f935d75049d78111d723fdce3780.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 196608 || token >= 229376) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 196608) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/59ea88d691b919af03bbb26f0253decfbc35d74f9661cdd3a08570ae85c313d2.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/59ea88d691b919af03bbb26f0253decfbc35d74f9661cdd3a08570ae85c313d2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5330d07c480a06ace52bbbeb6bf873fea1416d6c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cb30fce4d853f119d46ddb9bb88a134544f6362c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5a2219fcfa3b58bcbc86ec5535630f9dbe929a28999b5384c8eb10ba7b944684.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 1536u) % 1u) * 2048u + ((i / 768u) % 2u) * 1024u + ((i / 192u) % 4u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5abee175a331788afb292784154e8b92360ae6c22a818e351eaa15f5352f78ad.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5abee175a331788afb292784154e8b92360ae6c22a818e351eaa15f5352f78ad.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e8985ab18d2d8577f347f5cbefdea46af4bd69e5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5abee175a331788afb292784154e8b92360ae6c22a818e351eaa15f5352f78ad.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5c01813f337b4def8b4ed83afc43d6f55fe7c556739d4a23c5699a6da24e99fe.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5c01813f337b4def8b4ed83afc43d6f55fe7c556739d4a23c5699a6da24e99fe.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..41b58f2ffd5c2da2eb47696239a556c82c41fcf7 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5c01813f337b4def8b4ed83afc43d6f55fe7c556739d4a23c5699a6da24e99fe.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 512u) % 1u) * 2048u + ((i / 64u) % 8u) * 256u + ((i / 64u) % 1u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5fd6e068841f5c054392bef564909dbf086bae9e4030f3625d1143244a08cbbf.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/5fd6e068841f5c054392bef564909dbf086bae9e4030f3625d1143244a08cbbf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d60668fd46afd65337a293687e115ac26feb2dab --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8c75b69d7022cf5cd20457ed9ab0f9c3441f501e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6076d0bf36bd399e99fc409a95634cc3a508bec06ca558f0177434547983cd8b.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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] = f16(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f16(b1[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a2e5abc74d23b8a9d25fa3bc4097bb14e796338e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6292c78ff59552bb9d6a76d00232cc097e076bf7305d7b2539d3a2600b7d9f39.wgsl @@ -0,0 +1,18 @@ +enable f16; +@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 batch = i / 16384u; +let row = (i / 4096u) % 4u; +let col = i % 4096u; +var acc = 0.0; +for (var p = 0u; p < 256u; p++) { + acc += f32(b0[(((batch / 1u) % 8u) * 1024u) + row * 256u + p]) * f32(b1[(((batch / 1u) % 8u) * 1048576u) + p * 4096u + col]); +} +out[i] = f16(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/62f30a62c18ed78c9bf37bdc56529accbee2264b9eac931ba0b35a80bb917e7c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/62f30a62c18ed78c9bf37bdc56529accbee2264b9eac931ba0b35a80bb917e7c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f77c378d9485e77efa187d66dd4704a129bbf949 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..75fc18d394ecadc236f2b71f63ee719cec0cdd8e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/63207f8bf1aef72b1c386b9afcde2dd2885a6705ad0a7ae8223fc55259ec6483.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; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 1u) % 64u) * 128u) + (((j / 1u) % 128u) * 1u)]); } +out[i] = f32(acc / 128.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/64290f9bb656bb072a24effb35742fb11ced6f3bd4bc5f4b54123c9606cef41f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/64290f9bb656bb072a24effb35742fb11ced6f3bd4bc5f4b54123c9606cef41f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3edac89133b814ba21156e7f9e6b4307f7ed2041 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/64290f9bb656bb072a24effb35742fb11ced6f3bd4bc5f4b54123c9606cef41f.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/662fd2fc46677d76793ddad48cba911650d1b1fd722523cecbb5672ceed5abd5.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/662fd2fc46677d76793ddad48cba911650d1b1fd722523cecbb5672ceed5abd5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..47f909d40123ab7f14fbebb6254e445e691c94d4 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/662fd2fc46677d76793ddad48cba911650d1b1fd722523cecbb5672ceed5abd5.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 512u; + if (i >= 512u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 512u) * 2048u + p]) * f32(b1[(i % 512u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/665cda3c121e57eca4b72aa63e61e2e3d7935fb204d6fbc57adbb0b393fd32d4.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/665cda3c121e57eca4b72aa63e61e2e3d7935fb204d6fbc57adbb0b393fd32d4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..15171e78d3a57e3d50f4589f93af74fcc82fdd26 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/665cda3c121e57eca4b72aa63e61e2e3d7935fb204d6fbc57adbb0b393fd32d4.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 16u) % 1u) * 16u + ((i / 1u) % 16u) * 1u + ((i / 1u) % 1u) * 16u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/67374f90369a4c4b8fd88cdd3f6da26f1950cd7402ae7e7abb1c358bfe6cc0cd.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/67374f90369a4c4b8fd88cdd3f6da26f1950cd7402ae7e7abb1c358bfe6cc0cd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d37a092e9803acfc039bf2d5a35392bf84710644 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/67374f90369a4c4b8fd88cdd3f6da26f1950cd7402ae7e7abb1c358bfe6cc0cd.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 >= 1u) { return; } +var acc = 0.0; +for (var j = 0u; j < 2048u; j++) { acc += f32(b0[(((i / 1u) % 1u) * 2048u + ((i / 1u) % 1u) * 2048u) + (((j / 1u) % 2048u) * 1u)]); } +out[i] = f32(acc / 2048.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f985c7028aea2c721131c5bb33d01a9be7b8c79e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/679519988318785912522df0365b64c23c3249e7ca74b411ff3438740649d4f0.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b1[((i / 64u) % 4u) * 64u + ((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a337b88a5efde2d59ea676f0cfb58df2003935428bcc283950b7c89cab1aa7f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a337b88a5efde2d59ea676f0cfb58df2003935428bcc283950b7c89cab1aa7f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ae82bf1bbb32e6f2388f8a1ff2794088f359e940 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a337b88a5efde2d59ea676f0cfb58df2003935428bcc283950b7c89cab1aa7f.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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] = f16(b0[(i / 5u) * 4u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 4u && coord < 5u) { out[i] = f16(b1[(i / 5u) * 1u + (coord - 4u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a7b2e3778a02c76155d46fa84aeb33f594954e82e09b6b7f2915c0a606eb3f5.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a7b2e3778a02c76155d46fa84aeb33f594954e82e09b6b7f2915c0a606eb3f5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dbb8e11929641cd77e1054eb002c379556e71409 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..7f60d938f1501d684213feaaf57ce5a3313c853f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a83253d850e0909ae680acaa0774c5ad110f4c057980106c0383d73371bfb1a.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a953fa40ac66d71c6b96e0a14fb321f2d0e8c39a9af7041966bc07284720fa7.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a953fa40ac66d71c6b96e0a14fb321f2d0e8c39a9af7041966bc07284720fa7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fc79c9501c735aa0060d91e3691b02df98a24f56 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6a953fa40ac66d71c6b96e0a14fb321f2d0e8c39a9af7041966bc07284720fa7.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b1[((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6acb95aac27b4ee1945a1ea077b48448a1a91509a577269847d244a90b32c132.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6acb95aac27b4ee1945a1ea077b48448a1a91509a577269847d244a90b32c132.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..40b4a6fe0f957cedc4af44677a5a80f61e7f13ed --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6acb95aac27b4ee1945a1ea077b48448a1a91509a577269847d244a90b32c132.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 6144u + ((i / 2048u) % 1u) * 6144u + (((i / 1u) % 2048u) * 1u + 4096u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1fe897120a210af9b464dc043ead7462ce9138be --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6b66940e2070782344b4a34db496780b0080d9e6a1f51a9e31ca72e2e7b03790.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 / 2048u) % 4u) * 8u + ((i / 256u) % 8u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6be08553b45679d139ac09782dd180d640757cc4ed4f1d00ceeeb65dc470e72d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6be08553b45679d139ac09782dd180d640757cc4ed4f1d00ceeeb65dc470e72d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..31132550b1862e7d7dae65904ec250b533f8dbaa --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6be08553b45679d139ac09782dd180d640757cc4ed4f1d00ceeeb65dc470e72d.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 6144u + ((i / 2048u) % 1u) * 6144u + (((i / 1u) % 2048u) * 1u + 2048u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6dd75d41e19dbacfe48721ca98e0e9b910ff5ffdada05880789c0a011675fabf.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6dd75d41e19dbacfe48721ca98e0e9b910ff5ffdada05880789c0a011675fabf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c8beea40d7b77a9ea311e83ba103b73d67510f1b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/6dd75d41e19dbacfe48721ca98e0e9b910ff5ffdada05880789c0a011675fabf.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 >= 8u) { return; } +var acc = 0.0; +for (var j = 0u; j < 256u; j++) { acc += f32(b0[(((i / 8u) % 1u) * 2048u + ((i / 8u) % 1u) * 2048u + ((i / 1u) % 8u) * 256u) + (((j / 1u) % 256u) * 1u)]); } +out[i] = f32(acc / 256.0); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f9ada702dd9945723b880ef0b361ed4ebba55aaf --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/703d2f8d138453e2e9dc7709c73f9554630c33669e4c98c67997cbcd57ec9f81.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 8192u + ((i / 256u) % 8u) * 1024u + ((i / 64u) % 4u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..abe7259573449d6c991bafc90c853f251595383e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/70c92a5523efa88d08cf77af320f7506c59f501c993b093ae6472a55704606b2.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 2048u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 2048u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 2048u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/70ce424e071e4be502db10d94f3d121cbf8d15d920db399cc6e3a0727c060caa.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/70ce424e071e4be502db10d94f3d121cbf8d15d920db399cc6e3a0727c060caa.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9ffc77e52ab934575b1385965da0ee64b3e236d6 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/70ce424e071e4be502db10d94f3d121cbf8d15d920db399cc6e3a0727c060caa.wgsl @@ -0,0 +1,12 @@ +enable f16; +@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; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 16u) % 1u) * 2048u + ((i / 16u) % 1u) * 2048u + ((i / 1u) % 16u) * 128u) + (((j / 1u) % 128u) * 1u)]); } +out[i] = f16(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7121b5c223da0b0f28991f55cfed3c3f3390caa0bf65a8e58ae380db281ceb6b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7121b5c223da0b0f28991f55cfed3c3f3390caa0bf65a8e58ae380db281ceb6b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..69d777ae7caf63f2db83f47bb596dbe1e85a7c4b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7183997261247fa4df7e081c0d6d01036fee95b9b57ef944d5f44f72fc80d5fe.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7183997261247fa4df7e081c0d6d01036fee95b9b57ef944d5f44f72fc80d5fe.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..99d3e6e77543d2cc58b39e9dcb97109efd21a2fe --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7183997261247fa4df7e081c0d6d01036fee95b9b57ef944d5f44f72fc80d5fe.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]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e04fec04a49521c51d18b8903e46c9daf717004b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/73c884912738b1cca460f4ab298a530b8e6b543836ba47c0b507e010d8b2afba.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..51d32fb6edace0e5333396249372aee1a9034d87 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/73eb1bca0f997b4f6eaf7417b65024d79d49bed04a231daa562d2c252425ba6d.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 >= 32u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/742db031d819fdf5a756f37e8e824e2dcb2ca3f30efac3ffbd5918181059ae3a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/742db031d819fdf5a756f37e8e824e2dcb2ca3f30efac3ffbd5918181059ae3a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0879520c73bf4f1395e8a15cc5874fa97dbecde1 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/74825c737a9af6b5b5f135b1df88afaa7e80c3dfbb2ed9dd406e9242c50bfe06.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/74825c737a9af6b5b5f135b1df88afaa7e80c3dfbb2ed9dd406e9242c50bfe06.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8fb0a3ba28b97ae86dab2d3f04a56279c23362da --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/74825c737a9af6b5b5f135b1df88afaa7e80c3dfbb2ed9dd406e9242c50bfe06.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] = f32(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1601c2f1c95920dd739e32da0111d879da60655e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8a85c166513d8b4f82123beb8df8e37f8bbe085a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/759c63c4e237a2b3764b626e9db4f60536faf26577419d488b4fbf271a6e4e4a.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]) + (1.0 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d565ab36949bc42c6910f8cdf114a7d266e567a2 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7697257890a94501017b73d060c613444d6731c1191953c573305adcea662e3f.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0a6bcfd227464384d89d21677ed7dcd99dc3425e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/777b69ccbce41d844bbde50c84294eec98066ce09c0a49249c0083e6a98ab561.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f32(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7928d1b13d79c16f6c0f9d3130cbdb4193cab6efe3a60b38caa1039688fe56de.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7928d1b13d79c16f6c0f9d3130cbdb4193cab6efe3a60b38caa1039688fe56de.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d569f2c70babe1c72d9fa65e789ae8c5bdd2f048 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7ad635ba3bd72099f5b84597ad23ed54b83873ac53944649fd0bedb25a52a692.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7ad635ba3bd72099f5b84597ad23ed54b83873ac53944649fd0bedb25a52a692.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..db6e11433bc4264230e8dc7653c899197b935ac5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7ad635ba3bd72099f5b84597ad23ed54b83873ac53944649fd0bedb25a52a692.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 2048u + ((i / 128u) % 16u) * 128u + ((i / 128u) % 1u) * 2048u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..10b1bfcbdf17dfbfbbd6906648ef66fd0859c3c5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7ccffd09d8c42de21672bb1ded5c28f89737496c7696af68eb078dc71050191c.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 8192u) % 1u) * 24576u + ((i / 2048u) % 4u) * 6144u + (((i / 1u) % 2048u) * 1u + 4096u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4863306c4c96260e6a6b65fa59184f8a354fedb9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7dd89f4febc5d9accc7080e011d98765e6e31e55e8a44c5d1e384bc66dab5325.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7e90b8d0c7b998d4729541adc2de35c00fb0780d7ca3e89c66cb32638c3b533d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7e90b8d0c7b998d4729541adc2de35c00fb0780d7ca3e89c66cb32638c3b533d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5be6bc1a86c8a252a9913c57259b490b04f7b4c4 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7e90b8d0c7b998d4729541adc2de35c00fb0780d7ca3e89c66cb32638c3b533d.wgsl @@ -0,0 +1,23 @@ +enable f16; +@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 b6: array; +@group(0) @binding(7) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 917504u; +if (i >= 917504u) { return; } +let coord = (i / 1u) % 229376u; +if (coord >= 0u && coord < 32768u) { out[i] = f16(b0[(i / 229376u) * 32768u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32768u && coord < 65536u) { out[i] = f16(b1[(i / 229376u) * 32768u + (coord - 32768u) * 1u + i % 1u]); } +if (coord >= 65536u && coord < 98304u) { out[i] = f16(b2[(i / 229376u) * 32768u + (coord - 65536u) * 1u + i % 1u]); } +if (coord >= 98304u && coord < 131072u) { out[i] = f16(b3[(i / 229376u) * 32768u + (coord - 98304u) * 1u + i % 1u]); } +if (coord >= 131072u && coord < 163840u) { out[i] = f16(b4[(i / 229376u) * 32768u + (coord - 131072u) * 1u + i % 1u]); } +if (coord >= 163840u && coord < 196608u) { out[i] = f16(b5[(i / 229376u) * 32768u + (coord - 163840u) * 1u + i % 1u]); } +if (coord >= 196608u && coord < 229376u) { out[i] = f16(b6[(i / 229376u) * 32768u + (coord - 196608u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7fe183583b43b627721ba6853b306288d7ceea4b42b630410f47ea9637220a4b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/7fe183583b43b627721ba6853b306288d7ceea4b42b630410f47ea9637220a4b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..afbf92723660dbb4d364ef0cbeaf40d4cdc4891a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/82bc226410ee45662442e25040b6a6456c2532b87e96860e10f8f0135dde892b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/82bc226410ee45662442e25040b6a6456c2532b87e96860e10f8f0135dde892b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..28de6e0e30293bf23a7829da9811cd25682a84fa --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/82bc226410ee45662442e25040b6a6456c2532b87e96860e10f8f0135dde892b.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]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..df7347776d1e8bb5c2d9be3b7bc6488488e7e54e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8527da3511965e56b55949265779ba1a6a61b2257e7120ec97b4d4350cb2599c.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(-x); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/869b5ed8d1728d6fc7262ed6a0d8412dfd1e2c0ae420f068c41fc1f3f3ca14e9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/869b5ed8d1728d6fc7262ed6a0d8412dfd1e2c0ae420f068c41fc1f3f3ca14e9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..40d827d22e06007c30b9546cde391737d54a6f9e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/869b5ed8d1728d6fc7262ed6a0d8412dfd1e2c0ae420f068c41fc1f3f3ca14e9.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 * 512u; +if (i >= 512u) { return; } +let x = f32(b0[i]); +out[i] = f32(x*x); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b89d97c358c99704b0764e2b19561546e3e7208f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8a76ca278f842fa737a70b19b0676adcec489375a3a56de17ac9e354a687601e.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] = f32(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8bbc8f15f6eae5abd8e43cdadd577047e3e81811c701da1954f1355f3a514a21.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8bbc8f15f6eae5abd8e43cdadd577047e3e81811c701da1954f1355f3a514a21.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2ef17495943820b526b78ed8d3feb3e4b651dd33 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8bbc8f15f6eae5abd8e43cdadd577047e3e81811c701da1954f1355f3a514a21.wgsl @@ -0,0 +1,18 @@ +enable f16; +@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 batch = i / 256u; +let row = (i / 256u) % 1u; +let col = i % 256u; +var acc = 0.0; +for (var p = 0u; p < 4096u; p++) { + acc += f32(b0[(((batch / 1u) % 8u) * 4096u) + row * 4096u + p]) * f32(b1[(((batch / 1u) % 8u) * 1048576u) + p * 256u + col]); +} +out[i] = f16(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3d757c4e3617e29983738ee27efea1d4b36919cd --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8c34b573141052889a39d3eda528d755826e24ce48dc3e130e518e9be40a745b.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b1[((i / 2048u) % 4u) * 16u + ((i / 128u) % 16u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8cce7bfa491959f6a7b62ae9cb861d12e2f4ff05227918a4d06bd80669f0a539.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8cce7bfa491959f6a7b62ae9cb861d12e2f4ff05227918a4d06bd80669f0a539.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..afdc69dc467b8f7f632ffe33bddaf90b521e8362 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8a01c2daa51364f3125cbebcab9dbf7871446a9a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ded9f1437e5008402716aa09ca2c289948b0f9a0 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d5321128d8071a532e5cdf99bb40dad32139297770b61e746a9d8f0c9ff8af3.wgsl @@ -0,0 +1,18 @@ +enable f16; +@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 batch = i / 1024u; +let row = (i / 256u) % 4u; +let col = i % 256u; +var acc = 0.0; +for (var p = 0u; p < 4096u; p++) { + acc += f32(b0[(((batch / 1u) % 8u) * 16384u) + row * 4096u + p]) * f32(b1[(((batch / 1u) % 8u) * 1048576u) + p * 256u + col]); +} +out[i] = f16(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d9b846ecd9fe2c3acde76737571459a46136953e170abfca48b642fe9dce367.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d9b846ecd9fe2c3acde76737571459a46136953e170abfca48b642fe9dce367.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..448890d3d61498b7890b5c78aa3a08ac0c54a705 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8d9b846ecd9fe2c3acde76737571459a46136953e170abfca48b642fe9dce367.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 18944u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 18944u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 18944u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3de2aeb74e28396ce4fd20350a7ecfddd7fc01fb --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8f18ba35437ae0887d4d696e1494a47a5bd1a2a17befc070189db062616d6ece.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]) + (1.0 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..efe18f715642ac3e9e49928da7a2e1d17cd42b11 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/8fe40f3a29b0d9ad1eee9dfc3fa4cc6ccfc82ee74a6f681087fb852d812649a6.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 6144u) % 1u) * 8192u + ((i / 768u) % 8u) * 1024u + ((i / 192u) % 4u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/90d5a717885fcf0ac83d93bc97ccc20676b71b30aab9d4635494baee38ba88ce.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/90d5a717885fcf0ac83d93bc97ccc20676b71b30aab9d4635494baee38ba88ce.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..958b5bbdbf512591753bf03cc8d1b81ebf996171 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/90d5a717885fcf0ac83d93bc97ccc20676b71b30aab9d4635494baee38ba88ce.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 384u) % 1u) * 512u + ((i / 192u) % 2u) * 256u + ((i / 192u) % 1u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9144d827dae4c52b80abb829178a9ed901ce5dbc6702ad94b0b355e8eaad5ea5.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9144d827dae4c52b80abb829178a9ed901ce5dbc6702ad94b0b355e8eaad5ea5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0e196cbf9c5f57d582e8784874efe3d9797e27fe --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d89ad6cc3c9fdb690bd825e00999f4ce46769505 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/91bccb9cf3db8dea674e2c5f1865a39fdda37f3aac58fd62c4b81f9b7388c5b0.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/91fe8f8184a87eb1a52ef4ca8ef340b9b209789ceaa5b41a47f2cd240571ab70.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/91fe8f8184a87eb1a52ef4ca8ef340b9b209789ceaa5b41a47f2cd240571ab70.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d8e27bd1672b84788d86a0ef33bd3a3d2169ccb9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/91fe8f8184a87eb1a52ef4ca8ef340b9b209789ceaa5b41a47f2cd240571ab70.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(x / (1.0 + exp(-x))); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9249c507b0da3d89085fd5f8327b7d37fa1764281d06066cbc208d6d7e2ac137.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9249c507b0da3d89085fd5f8327b7d37fa1764281d06066cbc208d6d7e2ac137.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5cfea1afadf35d76994716add40ab444c6aba27a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9249c507b0da3d89085fd5f8327b7d37fa1764281d06066cbc208d6d7e2ac137.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 131072 || token >= 163840) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 131072) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/93120194556482afd2620e1987865bb53877263ccc980a055dfa748894585e8c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/93120194556482afd2620e1987865bb53877263ccc980a055dfa748894585e8c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f81c777c9bbc8b0632baafa5d4567d8391a4d91b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/93120194556482afd2620e1987865bb53877263ccc980a055dfa748894585e8c.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/947f5cc8b3f87e1a0d4d322d9d9e71e2ed16ce6e1f7c91118678c164dc1c8752.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/947f5cc8b3f87e1a0d4d322d9d9e71e2ed16ce6e1f7c91118678c164dc1c8752.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2bb5bd4f5d110df131987419e6018288811ea172 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/947f5cc8b3f87e1a0d4d322d9d9e71e2ed16ce6e1f7c91118678c164dc1c8752.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 >= 96u) { return; } +out[i] = f32(b0[((i / 32u) % 3u) * 32u + ((i / 32u) % 1u) * 32u + ((i / 1u) % 32u) * 1u + ((i / 1u) % 1u) * 32u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5c191c80ecf1a1e88202d99bac5c39f2638a2c4a --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/950c25980d19d89542840ceb582dfed03091d6651fc667d1340bb5ceee6bb8cb.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9c868b10c7c34316c00e41b14857af7bbfe43f60 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9625ca0b071511e5261877fd21627ca354ea0219da86a10887e8054afa15dcfd.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9625ca0b071511e5261877fd21627ca354ea0219da86a10887e8054afa15dcfd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e1f0cf834667ef3ff29080e64d8f9d08e6702c3e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cbb71ca117fff7c15242e53f62b62bf9bdf65ac4 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/97ad1988a80efddc69dea78e7053a235ebcf459153019ec7def6160064a5030b.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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] = f16(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f16(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/985bb9ad33654b9adbd4a256d626e346556a10dad514692800da8ebd7910f0f0.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/985bb9ad33654b9adbd4a256d626e346556a10dad514692800da8ebd7910f0f0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9018c3fdbf086b98282a94876e8a6631eb5442af --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/990406d9e76238db8bc85793325e969e48c04c75d1a217fe63b7cb7d9c561edf.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/990406d9e76238db8bc85793325e969e48c04c75d1a217fe63b7cb7d9c561edf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..18c672a602fa05263490c3742b7b25cc5e6e2a87 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/990406d9e76238db8bc85793325e969e48c04c75d1a217fe63b7cb7d9c561edf.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0cd8ab30fa23697b68e6ac3a9f5feadcc77b9d27 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/99e587565871f38abe78bb40bf99bd694d1cedd886da8565b2957ce110926c56.wgsl @@ -0,0 +1,22 @@ +enable f16; +@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] = f16(acc + 0.0); + +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/99fb5353b82923ddccb0cf31d2b470e745c8c3b98e1a5c9667ae323babe79b0a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/99fb5353b82923ddccb0cf31d2b470e745c8c3b98e1a5c9667ae323babe79b0a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..42eb59f6091c887537b1912618a5866953fe8f48 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/99fb5353b82923ddccb0cf31d2b470e745c8c3b98e1a5c9667ae323babe79b0a.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..71dcbb47847045bb8b499ccbd13001a8e4e02d80 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9cf9bb9c9852a2682cf47232223e6656a19d2893aecddfdeddbfe5d369aa811d.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 / 512u) % 4u) * 2u + ((i / 256u) % 2u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e13a281c20a618b227cdda747fdecaeca171af334742b83a80d2019ebb96cbd.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e13a281c20a618b227cdda747fdecaeca171af334742b83a80d2019ebb96cbd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a9735a9a58576e2ee3a6d2db456ca8ddf1f786ca --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e13a281c20a618b227cdda747fdecaeca171af334742b83a80d2019ebb96cbd.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 2048u) % 1u) * 6144u + ((i / 2048u) % 1u) * 6144u + (((i / 1u) % 2048u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6f477cffe09f8dd9cc82924861e9cf30629dbabc --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e19fc7e868893f5894ee5748adbd964074dae97f57a1bcb720a4e9421cecb22.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f16(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f16(b1[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e5386586f692cd1d2e6b25dce87abaa97386fa0ae759586e718dda142bea88a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e5386586f692cd1d2e6b25dce87abaa97386fa0ae759586e718dda142bea88a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5d7457ce6d9bfd674387b56096bdaa0ab9c31235 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/9e5386586f692cd1d2e6b25dce87abaa97386fa0ae759586e718dda142bea88a.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b1[i])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a051a8e7a9e6b7624e879ea355312b0517d92e476e10e73847a706e396e9549a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a051a8e7a9e6b7624e879ea355312b0517d92e476e10e73847a706e396e9549a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..59bd9565fd17f30659c838675a8bb82c7b9de551 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a051a8e7a9e6b7624e879ea355312b0517d92e476e10e73847a706e396e9549a.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 16u; + if (i >= 16u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 16u) * 2048u + p]) * f32(b1[(i % 16u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..569cf9c97c073ba43846aef98cc0eff29f4660b4 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a0f5527a8119db1e573ef20248655a5eecb61c6b975c7b9fb9fed2afbdc90f1a.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 8192u) % 1u) * 16384u + ((i / 2048u) % 4u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 256u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..740acb214f3c2958f1ef041c26966b2f9fd81b6b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a2561f5de51467cbba9b2f74e213693630ba6c086fb0a1b5b97315d37bfb6942.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a2bd65fc2709e093644d6cbf0c292b4d71520009a87bedcf4bca8fcb5f5c15b8.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a2bd65fc2709e093644d6cbf0c292b4d71520009a87bedcf4bca8fcb5f5c15b8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a3b7ba18c0aa6c84088bfae203f30b202cf1cedc --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a37dac611b27158f19aa5bd70c43217286e610e733cbd602b91666b4af42d3ef.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a37dac611b27158f19aa5bd70c43217286e610e733cbd602b91666b4af42d3ef.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c0eaa7ea6d730c603c8ce3193ea9e00a3ec74433 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a39942521aae0a7d7464b7ea4a1ca5319a3413097899bb118f3b58db9ec8e78d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a39942521aae0a7d7464b7ea4a1ca5319a3413097899bb118f3b58db9ec8e78d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a2c1a1b841654c0b5722295a846e673f95680185 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a39942521aae0a7d7464b7ea4a1ca5319a3413097899bb118f3b58db9ec8e78d.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 65536 || token >= 98304) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 65536) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a3e14309c38bb9a32a31e38d756485dddef949ff49d956a50908e2247c7c8abe.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a3e14309c38bb9a32a31e38d756485dddef949ff49d956a50908e2247c7c8abe.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..47e3fe5da0a5ea34120099d3e14c89df2d73b5db --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a3e14309c38bb9a32a31e38d756485dddef949ff49d956a50908e2247c7c8abe.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 512u) % 1u) * 512u + ((i / 256u) % 2u) * 256u + ((i / 256u) % 1u) * 512u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0556e53f06e6f7f9bb870fcd99defbe831f63ba1 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a429ab30ff95ce1b0f0388498d823ff9c043ad70b21f8bb609bbcf87cd28014e.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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 * 2097152u; +if (i >= 2097152u) { return; } +out[i] = f16(b0[((i) / 256u) % 4096u * 512u + ((i) / 1048576u) % 2u * 256u + ((i) / 1u) % 256u * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a4cf1224f0c913592843045b6b129bb92273bc021af13296127d1b57156e6f88.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a4cf1224f0c913592843045b6b129bb92273bc021af13296127d1b57156e6f88.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..18d7749e6e4110f3707919a9875b0059f5ae1e2f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a4cf1224f0c913592843045b6b129bb92273bc021af13296127d1b57156e6f88.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b1[((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a58e9adbad973ba2f87419a87d15a4b203e4fac2adb1c2f0f45dded8467e7386.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a58e9adbad973ba2f87419a87d15a4b203e4fac2adb1c2f0f45dded8467e7386.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8dafafcc769f0a0153ea36a8a3d0e469a65fb915 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a58e9adbad973ba2f87419a87d15a4b203e4fac2adb1c2f0f45dded8467e7386.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 / 256u) % 8u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a672c3e158716bde7444c8141f3df1da16da739ccdd2eb2ad60f9d42d0775157.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a672c3e158716bde7444c8141f3df1da16da739ccdd2eb2ad60f9d42d0775157.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..aa456f658303b9267eb9faab7425cd740bbdc6f5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/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]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8f7e080e8cee8c2acabacf51697ca01fe402a237 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.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 >= 96u) { return; } +out[i] = f32(b0[((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a87152b0fd2f78f60d7abb61c2f6c4a9bde8a9dd45d0275b58d760a3da9f71ce.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a87152b0fd2f78f60d7abb61c2f6c4a9bde8a9dd45d0275b58d760a3da9f71ce.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a411756a9239877ab739c80010eb78b97e1cc130 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/a87152b0fd2f78f60d7abb61c2f6c4a9bde8a9dd45d0275b58d760a3da9f71ce.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 * 384u; +if (i >= 384u) { return; } +let batch = i / 128u; +let row = (i / 4u) % 32u; +let col = i % 4u; +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) * 4u) + p * 4u + col]); +} +out[i] = f32(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/accc4c502a4ddeea7dc9c3d00f3f58cd77ecb5ccddc0f1dd44a2c578e97e63f2.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/accc4c502a4ddeea7dc9c3d00f3f58cd77ecb5ccddc0f1dd44a2c578e97e63f2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cea97ffa48d7d65929a78eb9126ca5158470292d --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/accc4c502a4ddeea7dc9c3d00f3f58cd77ecb5ccddc0f1dd44a2c578e97e63f2.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ae69e698c92571766e36ad8c078b13e42bde56745d45698daaadaca587b7279c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ae69e698c92571766e36ad8c078b13e42bde56745d45698daaadaca587b7279c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..48686d13bb0682a3b49a3239768a42f4ae4217d8 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ae69e698c92571766e36ad8c078b13e42bde56745d45698daaadaca587b7279c.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 >= 2u && coord < 32u && (coord - 2u) % 3u == 0u) { out[i] = f32(b0[((i / 128u) % 1u) * 40u + ((i / 32u) % 4u) * 10u + ((((i / 1u) % 32u) - 2u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b18c5b1cd31ea00868fbfeecb448516b406bf6bf --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ae8f63dd06507fd0c575e576c126944275a62fd9b6730313e8e5ac6369872267.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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 * 4194240u; +if (i >= 8388608u) { return; } +out[i] = f16(b0[((i / 8388608u) % 1u) * 8388608u + ((i / 1048576u) % 8u) * 1048576u + ((i / 4096u) % 256u) * 1u + ((i / 1u) % 4096u) * 256u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b8c5b5973ce2dc23fee65ff3aa50f18b4c08b682 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/af8bae1f95fe663a390466c647dc794083b2250ae839785f57fb2c7483f3b3ec.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 49152u) % 1u) * 67584u + ((i / 8u) % 6144u) * 11u + (((i / 1u) % 8u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..add0a490bc8865ca3ed0caa4e0d09c317d9c6917 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/afac69942bbaa6f0b7e1ff71a4ddc7ea264cbdb4a080d75b9a9856fdc12554b6.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 >= 4u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a1b0e550d8c9fafe0d2d025de16ace6169cc4cff --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b0838d1f95f0fd6641bd3ae7d989ebe8f964bc3c82631b2a33e2ca564a927e9b.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fb7196f3862e62c9193025db5acc14ee1544e4ac --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b0ed4ebf2fb21d2418ea5c98632a10939152ab3eb2394ac6389675ec79b5d675.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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] = f16(b0[(i / 8u) * 4u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 4u && coord < 8u) { out[i] = f16(b1[(i / 8u) * 4u + (coord - 4u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b51b408468094333d9becb8bad900c51d320e3e2 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b156838d80bcdf65a474c8fb509840619fe2f81849005c053cfd5bfdd714a8a4.wgsl @@ -0,0 +1,12 @@ +enable f16; +@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; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 64u) % 1u) * 8192u + ((i / 16u) % 4u) * 2048u + ((i / 1u) % 16u) * 128u) + (((j / 1u) % 128u) * 1u)]); } +out[i] = f16(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b17aaca06f3b631ebb0a4e873183966913175acb34bb0cec516c8bbec20c22dd.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b17aaca06f3b631ebb0a4e873183966913175acb34bb0cec516c8bbec20c22dd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..96c11ce5bd92296c59a188e7ec61fe4d80950ca8 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b17aaca06f3b631ebb0a4e873183966913175acb34bb0cec516c8bbec20c22dd.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 65536 || token >= 98304) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 65536) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b199d835db70f888c7b99a3ac6e9ec56f6c47d5bba4eaf352e0a8eabed56411e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b199d835db70f888c7b99a3ac6e9ec56f6c47d5bba4eaf352e0a8eabed56411e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..54526e84e0f57f0b27c2e0d57273b25cde83790f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b199d835db70f888c7b99a3ac6e9ec56f6c47d5bba4eaf352e0a8eabed56411e.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 / 128u) % 16u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9b18092deac8f9ecf759c438a10098e2899a4448 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b28ea4a8345a9c9abb2d74bfd620ba5de46616434dbaffdbabf022ea71800839.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]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9acdd6e6de3ab6228cb265912c8aa1754f86ee70 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b5a8f92a74987b61c9ec5720a8a90a4d36b633a2f71e9233d8d691a1bc41e20e.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 24576u) % 1u) * 24576u + ((i / 4u) % 6144u) * 1u + ((i / 1u) % 4u) * 6144u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..da7a62945e0fcc1ae5d2cfacf4ab0deb299416a5 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b6d8ca3438a5d12cca03c5223ed58ff381d16e43b4eb1c9a78ca1a49222e9b8e.wgsl @@ -0,0 +1,32 @@ +enable f16; +@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 * 256u + col / 8u]; + let code = i32((word >> ((col % 8u) * 4u)) & 15u) - 8; + return f32(f16(f32(code) * b2[row * 256u + col / 8u])); +} + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 6144u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 6144u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = dequant_1((group.x * 8u + local.y) * 2048u + tile + local.x); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 6144u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b7a4fe3a16a37ec4eb65ad6959dc83f30aa20cdc405aa66ee0d4efeb1920b1a1.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b7a4fe3a16a37ec4eb65ad6959dc83f30aa20cdc405aa66ee0d4efeb1920b1a1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1a8022d0caa67680e866df6ba6fd74e6bad97552 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b7a4fe3a16a37ec4eb65ad6959dc83f30aa20cdc405aa66ee0d4efeb1920b1a1.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 128u) % 1u) * 512u + ((i / 64u) % 2u) * 256u + ((i / 64u) % 1u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b89e318eaf367e026643645379e68b177a3b2054 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/b878d3aa797f91de4d478205469ddf14f86faa394e0473cdb02f745fc472e5fe.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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; } +out[i] = f16(f32(b0[i]) + (f32(b1[((i / 4096u) % 4u) * 4096u + ((i / 1u) % 4096u) * 1u]) * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ba2b60fc7e0bf8f703260ce0eb714460e8411fe09096a0d1950298537a0b0d3a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ba2b60fc7e0bf8f703260ce0eb714460e8411fe09096a0d1950298537a0b0d3a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ac5a2a85189aa1e502e893e122416de1e51ce842 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ba2b60fc7e0bf8f703260ce0eb714460e8411fe09096a0d1950298537a0b0d3a.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 < 229376u) { out[i] = f16(b0[(i / 248320u) * 229376u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 229376u && coord < 248320u) { out[i] = f16(b1[(i / 248320u) * 18944u + (coord - 229376u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bb07991ad7088f42c159cbd086520fb2900c45cf27ed91784f2dc1780d36ed70.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bb07991ad7088f42c159cbd086520fb2900c45cf27ed91784f2dc1780d36ed70.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1cf16693c6d5f13164839ce21ecde410dc81145c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bb07991ad7088f42c159cbd086520fb2900c45cf27ed91784f2dc1780d36ed70.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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; } +out[i] = f16(b0[((i / 30720u) % 1u) * 49152u + ((i / 5u) % 6144u) * 8u + (((i / 1u) % 5u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bb9029b667eeb13670d1f640cd677c6eabd7a7569686c599e3eeb9482d642b00.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bb9029b667eeb13670d1f640cd677c6eabd7a7569686c599e3eeb9482d642b00.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..70d42d4c763bcca299e524c1695c1ae352353cd9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bb9029b667eeb13670d1f640cd677c6eabd7a7569686c599e3eeb9482d642b00.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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bbe6ba91174ed6aaebef3999a11f7bcd273153e6fa8936de2f884e9d53ab0857.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bbe6ba91174ed6aaebef3999a11f7bcd273153e6fa8936de2f884e9d53ab0857.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..64a9397bddd42503502dbf1de7d6430acbff5b91 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bbe6ba91174ed6aaebef3999a11f7bcd273153e6fa8936de2f884e9d53ab0857.wgsl @@ -0,0 +1,29 @@ +enable f16; +@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 * 256u + col / 8u]; + let code = i32((word >> ((col % 8u) * 4u)) & 15u) - 8; + return f32(f16(f32(code) * b2[row * 256u + col / 8u])); +} + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 6144u; + if (i >= 6144u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 6144u) * 2048u + p]) * dequant_1((i % 6144u) * 2048u + p); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fa0957030a3e0cec5ec76a848a5251574b4cd57c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bc279fe3036296f06a42b850ab0ffa35ce8f3543de5806fbb1da45291fba0695.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 1024u) % 1u) * 2048u + ((i / 128u) % 8u) * 256u + ((i / 32u) % 4u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..038f9a0691d60ac6354f403f7b365492967e2b35 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bcbd50e0ebcb6af3745d01bbe0f67cc268568bc4099595f36eea826e94b19e28.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 512u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 512u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 512u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bcbe14ef2fceea3448f416019a200c774bbd023457d9c833f20bf6c674bfa87f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bcbe14ef2fceea3448f416019a200c774bbd023457d9c833f20bf6c674bfa87f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..459154deb1b86e2c50aef0bec16e5b41ad797c90 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bcbe14ef2fceea3448f416019a200c774bbd023457d9c833f20bf6c674bfa87f.wgsl @@ -0,0 +1,22 @@ +enable f16; +@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] = f16(acc + 0.0); + +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9ebd7e231f4fdbcfd34fa3b1f3cab07ca3c2bb16 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bf4e2f4388f550b7ae4cb7cc6c7a2d9f9a03586475eb2d260440b00702c211f7.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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] = f16(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b5477a9f8d229c3fecaea22657917c108b99f833 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/bfe28ab64ac790ec085608fee6e3a68db22b0136a0a5b225677b1ef70d761641.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 24576u) % 1u) * 49152u + ((i / 4u) % 6144u) * 8u + (((i / 1u) % 4u) * 1u + 4u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c0b2947798baa198598d75ca0e5dcd31a9fa320499665e34f7bb81d27fc15fb0.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c0b2947798baa198598d75ca0e5dcd31a9fa320499665e34f7bb81d27fc15fb0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..bb8b3676ab0dfbc0c5c78d28f96387bc3a721046 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c0b2947798baa198598d75ca0e5dcd31a9fa320499665e34f7bb81d27fc15fb0.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(cos(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c23bee3c3664172bf0a078a96acd2b3ad3136732234975f757d313e696c2c37a.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c23bee3c3664172bf0a078a96acd2b3ad3136732234975f757d313e696c2c37a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2d80755364a4fee024c589b10f9f00932c251576 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c23bee3c3664172bf0a078a96acd2b3ad3136732234975f757d313e696c2c37a.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 / 1u) % 16u) * 1u]) * f32(b1[i])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c2c806bbe93f09247867b4ace45489a5f2fe3ee67f265418eb7aa71a29880276.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c2c806bbe93f09247867b4ace45489a5f2fe3ee67f265418eb7aa71a29880276.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fa4df64a06b41dc3305ec748ea504d435a19aa23 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c2c806bbe93f09247867b4ace45489a5f2fe3ee67f265418eb7aa71a29880276.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(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c4e65292c406754db5cde9209846773aec380a2963ba6ca329ea68fe501ed3e0.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c4e65292c406754db5cde9209846773aec380a2963ba6ca329ea68fe501ed3e0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b63f152d20bd494cccff76d97fca2dc68d0faf04 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c4e65292c406754db5cde9209846773aec380a2963ba6ca329ea68fe501ed3e0.wgsl @@ -0,0 +1,11 @@ +enable f16; +@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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c5381a1922110bc02e83d499e6af952b8ee49786c658a3666896584b9d89a365.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c5381a1922110bc02e83d499e6af952b8ee49786c658a3666896584b9d89a365.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9c4c15cad8fa4ac4ac17abf649870cf29b5ca764 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c5381a1922110bc02e83d499e6af952b8ee49786c658a3666896584b9d89a365.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 163840 || token >= 196608) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 163840) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c5b8ff2b58dcb0588eaedfd65be9de8cdd53dc0f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c6e2017c1d1d988837ac987ad5fba898224f7b5f249bd0786d2eeaa8d809ef7d.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(f32(b0[i]) * f32(b0[i])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..461d50f61a4ef44be0b7751a3fcc151b94921a87 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c6e66b53a1081cbfe07ed7cf947c26a7a4531765b4d51ae93977befb02611a8f.wgsl @@ -0,0 +1,12 @@ +enable f16; +@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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c72041e82befee6919af48775291256e22a43dc6445ef0aa288db6b96585a2a9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c72041e82befee6919af48775291256e22a43dc6445ef0aa288db6b96585a2a9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ceffcce3113d363ae38d0760c9af3a93e2a924d3 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c72041e82befee6919af48775291256e22a43dc6445ef0aa288db6b96585a2a9.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[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f327dab2dcc347a7035d0ddb35b9f03e1f183d48 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c80a8e9a7d08644037dcc89d0faea51c11125e9212f09dac7a4fc32d1d4f2743.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c82dd7f7bd4afd0cf1031cecb00d8978b4be6c78f7349b0f2a78622d533bd432.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c82dd7f7bd4afd0cf1031cecb00d8978b4be6c78f7349b0f2a78622d533bd432.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..db7f91aa63d60510b9ee579f042a4ca4f4d413e9 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c82dd7f7bd4afd0cf1031cecb00d8978b4be6c78f7349b0f2a78622d533bd432.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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] = f16(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f16(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c8be6fb8617061857df35e7ab1603e5550b9ce7a31d73d75d77bed7cdc5d241d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c8be6fb8617061857df35e7ab1603e5550b9ce7a31d73d75d77bed7cdc5d241d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5538d477713e9208e8723f9dde9f1356af4f51e4 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c8be6fb8617061857df35e7ab1603e5550b9ce7a31d73d75d77bed7cdc5d241d.wgsl @@ -0,0 +1,13 @@ +enable f16; +@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 = i32(b0[i / 2048u]); +if (token < 32768 || token >= 65536) { out[i] = f16(0.0); return; } +out[i] = f16(b1[u32(token - 32768) * 2048u + i % 2048u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c90428425f6d220b441e97e3b894adfb090869c98b17f53afbd2f9fb202b4525.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c90428425f6d220b441e97e3b894adfb090869c98b17f53afbd2f9fb202b4525.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6bbc472bea9c8d96ae70e94b1ca3e146cd081981 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c90428425f6d220b441e97e3b894adfb090869c98b17f53afbd2f9fb202b4525.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(sin(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1c649782846f9e1dcf6968ebd567f37ebe8b46df --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/c9f8b71df7455ffe929e11acc2c75e85cf70b0857e8b2af97fa331cb30c7bb69.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 1024u) % 1u) * 2048u + ((i / 128u) % 8u) * 256u + ((i / 32u) % 4u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..76c282a5b12b3774fec62b3047b2e2e1403b2163 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cba4b788ea98e4b44a1564d36d3a471684264a132691ffaad374fc1a94f66a6c.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cbe7ec56ce7efa22fb07f71b374a4e8953e9dcff5c95fb000f6bf5295bbae63d.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cbe7ec56ce7efa22fb07f71b374a4e8953e9dcff5c95fb000f6bf5295bbae63d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9becf8c3f7218c82aeb4fcbe90ffe8f5efe85874 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cbe7ec56ce7efa22fb07f71b374a4e8953e9dcff5c95fb000f6bf5295bbae63d.wgsl @@ -0,0 +1,35 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 8u; + if (row >= 8u) { return; } + let lane = local.x; + var maximum = -3.4028234663852886e38; + for (var p = lane; p < 4096u; p += 64u) { maximum = max(maximum, f32(b0[row * 4096u + p])); } + partial[lane] = maximum; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] = max(partial[lane], partial[lane + stride]); } + workgroupBarrier(); + } + maximum = partial[0]; + // All lanes must finish reading the maximum before the scratch array is reused. + workgroupBarrier(); + var total = 0.0; + for (var p = lane; p < 4096u; p += 64u) { total += exp(f32(b0[row * 4096u + p]) - maximum); } + partial[lane] = total; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + total = partial[0]; + for (var p = lane; p < 4096u; p += 64u) { + let i = row * 4096u + p; + out[i] = f32(exp(f32(b0[row * 4096u + p]) - maximum) / total); + } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cc250879ffbf43ba05860b5a86f6a824e8ca7b4c9f820137a68e1ba633e09738.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cc250879ffbf43ba05860b5a86f6a824e8ca7b4c9f820137a68e1ba633e09738.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f99a31a5530ee04c3f3573c324dd7c6471dc1049 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/cc250879ffbf43ba05860b5a86f6a824e8ca7b4c9f820137a68e1ba633e09738.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 * 128u; +if (i >= 96u) { return; } +let batch = i / 32u; +let row = (i / 1u) % 32u; +let col = i % 1u; +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) * 1u) + p * 1u + col]); +} +out[i] = f32(acc); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ce21e77f8c5ec99a64e195c9e92fdde9ae295eeff969b4f3b26b79f565a8e25c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ce21e77f8c5ec99a64e195c9e92fdde9ae295eeff969b4f3b26b79f565a8e25c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ad1ca504c4085e397ea7590a93cb0c3fc487e193 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ce21e77f8c5ec99a64e195c9e92fdde9ae295eeff969b4f3b26b79f565a8e25c.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 6144u) % 1u) * 30720u + ((i / 1u) % 6144u) * 5u + (((i / 1u) % 1u) * 1u + 4u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..96ee54af219ec72436f759419395f9a0515a0f29 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/ce7f5578fa59e549ccce4f144cb5ab1b6dcdb9dcdd5d41f8c508035c348155e3.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d0218f4c7661846abccb6d0d4ce847d2d96466c9255f3a476d705bbf3dc8154c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d0218f4c7661846abccb6d0d4ce847d2d96466c9255f3a476d705bbf3dc8154c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..427188daa3124937ccfd3987c36f94fa80272f8b --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d0218f4c7661846abccb6d0d4ce847d2d96466c9255f3a476d705bbf3dc8154c.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) % 256u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f71ddd63d1119c6e879ca95cbd2478d0514d7ced --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d41487b3cfbaaabeb3498fef3fecb17a748aeb0800f53030c844e586459a9a2f.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(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..720ece912936b614b52a077c7fc0a05ff6191231 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d46c2e705dcff0ad6dc6b1861a83aa89f0936dc6a1df3d70905a2572a28cf5ba.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 / 128u) % 64u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..87e447d12560fbe319e6ffdf4cdc9ef3ac811adc --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d58d61d4a27263601b290179509857c9837633bac7e76a9e7a9fee9685d07277.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 16u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 16u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 16u) { out[i] = f16(acc + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d7328b7ea65acf83ac644f94eb36917955ca66392c0974e4198d0fb0477c7a78.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d7328b7ea65acf83ac644f94eb36917955ca66392c0974e4198d0fb0477c7a78.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3f7bbc0be7b396061c23aea86a6924fd4fccc7a2 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d7328b7ea65acf83ac644f94eb36917955ca66392c0974e4198d0fb0477c7a78.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[i]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d8c79a22e83e4a3b75b732bffa9ca673713bb57d580c5af4e59c805645ff1ca3.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d8c79a22e83e4a3b75b732bffa9ca673713bb57d580c5af4e59c805645ff1ca3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2b2fd35974b2d0fe841a8a268a0226630737a60e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d8c79a22e83e4a3b75b732bffa9ca673713bb57d580c5af4e59c805645ff1ca3.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(cos(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d8f1ce89a98a35a4f6161e762fe8f967a562dd64c6e26d82afca490c0287a0b9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d8f1ce89a98a35a4f6161e762fe8f967a562dd64c6e26d82afca490c0287a0b9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..29a959d8090cfc40f1f18718b27bf0e6d945426e --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/d8f1ce89a98a35a4f6161e762fe8f967a562dd64c6e26d82afca490c0287a0b9.wgsl @@ -0,0 +1,23 @@ +enable f16; +@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 b6: array; +@group(0) @binding(7) 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 coord = (i / 1u) % 229376u; +if (coord >= 0u && coord < 32768u) { out[i] = f16(b0[(i / 229376u) * 32768u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32768u && coord < 65536u) { out[i] = f16(b1[(i / 229376u) * 32768u + (coord - 32768u) * 1u + i % 1u]); } +if (coord >= 65536u && coord < 98304u) { out[i] = f16(b2[(i / 229376u) * 32768u + (coord - 65536u) * 1u + i % 1u]); } +if (coord >= 98304u && coord < 131072u) { out[i] = f16(b3[(i / 229376u) * 32768u + (coord - 98304u) * 1u + i % 1u]); } +if (coord >= 131072u && coord < 163840u) { out[i] = f16(b4[(i / 229376u) * 32768u + (coord - 131072u) * 1u + i % 1u]); } +if (coord >= 163840u && coord < 196608u) { out[i] = f16(b5[(i / 229376u) * 32768u + (coord - 163840u) * 1u + i % 1u]); } +if (coord >= 196608u && coord < 229376u) { out[i] = f16(b6[(i / 229376u) * 32768u + (coord - 196608u) * 1u + i % 1u]); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cfd1c44a95f375e21dcd10457e73b20acb8e6a8c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dba1f878f575cac1bab071f2c650c8a596025dfbfbab50746f53a142255df962.wgsl @@ -0,0 +1,10 @@ +enable f16; +@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] = f16(b0[((i / 8192u) % 1u) * 24576u + ((i / 2048u) % 4u) * 6144u + (((i / 1u) % 2048u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dc4ee648b97b273931e31c95b910d66f11e34a1cb036589e9ae9cc250b6e6291.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dc4ee648b97b273931e31c95b910d66f11e34a1cb036589e9ae9cc250b6e6291.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0e13dfcb4085c03a943bab386ba2afe5e2a6133c --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dc4ee648b97b273931e31c95b910d66f11e34a1cb036589e9ae9cc250b6e6291.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[0u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9aaeb8d02a99dc7022c3d0e38380a68009f5327f --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/dd4cf611f51e105c96375dfcdc8420e9c8c8cea23f0a4236bacacbec6bd92f20.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 / 1u) % 256u) * 1u])); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4f2bf7e4c7efd7610ae86a151b04332e7a137cb4 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.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(inverseSqrt(x)); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/deda4d3032222f1f1bd8cf059fa8131f8046be80fa046f227547f667d8f06e0c.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/deda4d3032222f1f1bd8cf059fa8131f8046be80fa046f227547f667d8f06e0c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2474c91c129be1c6c7267ef1c84e1a5185abf777 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/deda4d3032222f1f1bd8cf059fa8131f8046be80fa046f227547f667d8f06e0c.wgsl @@ -0,0 +1,21 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let i = group.x + group.y * 6144u; + if (i >= 6144u) { return; } + let lane = local.x; + var acc = 0.0; + for (var p = lane; p < 2048u; p += 64u) { acc += f32(b0[(i / 6144u) * 2048u + p]) * f32(b1[(i % 6144u) * 2048u + p]); } + partial[lane] = acc; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[i] = f16(partial[0] + 0.0); } +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/deee894ecb86334ff82870260fc5edbc36439c1d3bce09fab5708c03d699b390.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/deee894ecb86334ff82870260fc5edbc36439c1d3bce09fab5708c03d699b390.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..edaf6eca29e64fa2e4a563af4eaf9fe76bfb99b7 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/deee894ecb86334ff82870260fc5edbc36439c1d3bce09fab5708c03d699b390.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[2u * 32u + ((i / 32u) % 1u) * 32u + ((i / 32u) % 1u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3f32512b09a84a8dd38682c787812e7a54477088 --- /dev/null +++ b/qwen35-2b-fp16-int4-g8-gptq-up-proj-candidate/kernels/e3435949f53d19f2a15b6939246967709046cac57899e1d048df5e202654622b.wgsl @@ -0,0 +1,24 @@ +enable f16; +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var tile_a: array; +var tile_b: array; +@compute @workgroup_size(8, 8) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.y * 8u + local.y; + let col = group.x * 8u + local.x; + let i = (group.z * 4u + row) * 4096u + col; + var acc = 0.0; + for (var tile = 0u; tile < 2048u; tile += 8u) { + tile_a[local.y * 8u + local.x] = 0.0; + tile_b[local.x * 8u + local.y] = 0.0; + if (row < 4u && tile + local.x < 2048u) { tile_a[local.y * 8u + local.x] = f32(b0[(group.z * 4u + row) * 2048u + tile + local.x]); } + if (group.x * 8u + local.y < 4096u && tile + local.x < 2048u) { tile_b[local.x * 8u + local.y] = f32(b1[(group.x * 8u + local.y) * 2048u + tile + local.x]); } + workgroupBarrier(); + for (var p = 0u; p < 8u; p++) { acc += tile_a[local.y * 8u + p] * tile_b[p * 8u + local.x]; } + workgroupBarrier(); + } + if (row < 4u && col < 4096u) { out[i] = f16(acc + 0.0); } +}