Upload config.json with huggingface_hub
Browse files- config.json +34 -20
config.json
CHANGED
|
@@ -4,32 +4,17 @@
|
|
| 4 |
],
|
| 5 |
"image_token_id": 248056,
|
| 6 |
"model_type": "qwen3_5",
|
|
|
|
| 7 |
"tie_word_embeddings": false,
|
| 8 |
"transformers_version": "4.57.0.dev0",
|
| 9 |
-
"video_token_id": 248057,
|
| 10 |
-
"vision_config": {
|
| 11 |
-
"deepstack_visual_indexes": [],
|
| 12 |
-
"depth": 27,
|
| 13 |
-
"hidden_act": "gelu_pytorch_tanh",
|
| 14 |
-
"hidden_size": 1152,
|
| 15 |
-
"in_channels": 3,
|
| 16 |
-
"initializer_range": 0.02,
|
| 17 |
-
"intermediate_size": 4304,
|
| 18 |
-
"model_type": "qwen3_5",
|
| 19 |
-
"num_heads": 16,
|
| 20 |
-
"num_position_embeddings": 2304,
|
| 21 |
-
"out_hidden_size": 4096,
|
| 22 |
-
"patch_size": 16,
|
| 23 |
-
"spatial_merge_size": 2,
|
| 24 |
-
"temporal_patch_size": 2
|
| 25 |
-
},
|
| 26 |
"vision_end_token_id": 248054,
|
| 27 |
"vision_start_token_id": 248053,
|
|
|
|
|
|
|
| 28 |
"attention_bias": false,
|
| 29 |
"attention_dropout": 0.0,
|
| 30 |
"attn_output_gate": true,
|
| 31 |
"dtype": "bfloat16",
|
| 32 |
-
"eos_token_id": 248044,
|
| 33 |
"full_attention_interval": 4,
|
| 34 |
"head_dim": 256,
|
| 35 |
"hidden_act": "silu",
|
|
@@ -70,6 +55,34 @@
|
|
| 70 |
"linear_attention",
|
| 71 |
"full_attention"
|
| 72 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
"linear_conv_kernel_dim": 4,
|
| 74 |
"linear_key_head_dim": 128,
|
| 75 |
"linear_num_key_heads": 16,
|
|
@@ -97,5 +110,6 @@
|
|
| 97 |
"rope_theta": 10000000,
|
| 98 |
"partial_rotary_factor": 0.25
|
| 99 |
},
|
| 100 |
-
"
|
| 101 |
-
|
|
|
|
|
|
| 4 |
],
|
| 5 |
"image_token_id": 248056,
|
| 6 |
"model_type": "qwen3_5",
|
| 7 |
+
"model_type": "qwen3_5_text",
|
| 8 |
"tie_word_embeddings": false,
|
| 9 |
"transformers_version": "4.57.0.dev0",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
"vision_end_token_id": 248054,
|
| 11 |
"vision_start_token_id": 248053,
|
| 12 |
+
"eos_token_id": 248046,
|
| 13 |
+
"pad_token_id": 248044,
|
| 14 |
"attention_bias": false,
|
| 15 |
"attention_dropout": 0.0,
|
| 16 |
"attn_output_gate": true,
|
| 17 |
"dtype": "bfloat16",
|
|
|
|
| 18 |
"full_attention_interval": 4,
|
| 19 |
"head_dim": 256,
|
| 20 |
"hidden_act": "silu",
|
|
|
|
| 55 |
"linear_attention",
|
| 56 |
"full_attention"
|
| 57 |
],
|
| 58 |
+
"vision_model_weight": "vision_weight.q4nx",
|
| 59 |
+
"vision_end_token_id": 248054,
|
| 60 |
+
"vision_start_token_id": 248053,
|
| 61 |
+
"vision_config": {
|
| 62 |
+
"vision_mm_engine_xclbin_name": "vision_mm.xclbin",
|
| 63 |
+
"vision_mha_engine_xclbin_name": "vision_attn.xclbin",
|
| 64 |
+
"QWEN3_5_PATCH_SIZE": 16,
|
| 65 |
+
"QWEN3_5_IMAGE_MERGE_SIZE": 2,
|
| 66 |
+
"QWEN3_5_SPATIAL_MERGE_SIZE": 2,
|
| 67 |
+
"QWEN3_5_SHORTEST_EDGE": 65536,
|
| 68 |
+
"QWEN3_5_LONGEST_EDGE": 16777216,
|
| 69 |
+
"QWEN3_5_VISION_RESCALE_FACTOR": 0.00392156862745098,
|
| 70 |
+
"QWEN3_5_VISION_RESCALE_IMAGE_MEAN": 0.5,
|
| 71 |
+
"QWEN3_5_VISION_RESCALE_IMAGE_STD": 0.5,
|
| 72 |
+
"QWEN3_5_TEMPORAL_PATCH_SIZE": 2,
|
| 73 |
+
|
| 74 |
+
"QWEN3_5_VISION_EMBED_DIM": 1152,
|
| 75 |
+
"QWEN3_5_VISION_NUM_HEADS": 16,
|
| 76 |
+
"QWEN3_5_VISION_HEAD_DIM": 72,
|
| 77 |
+
"QWEN3_5_VISION_MLP_INTERMEDIATE_SIZE": 4304,
|
| 78 |
+
"QWEN3_5_VISION_NUM_POSITION_EMBEDDINGS": 2304,
|
| 79 |
+
"QWEN3_5_VISION_NUM_LAYERS": 27,
|
| 80 |
+
"QWEN3_5_VISION_LAYER_NORM_EPSILON": 1e-06,
|
| 81 |
+
"QWEN3_5_VISION_OUT_HIDDEN_SIZE": 4096,
|
| 82 |
+
"VISION_MM_TILE_M": 128,
|
| 83 |
+
"VISION_MM_TILE_K": 256,
|
| 84 |
+
"VISION_MM_TILE_N": 64
|
| 85 |
+
},
|
| 86 |
"linear_conv_kernel_dim": 4,
|
| 87 |
"linear_key_head_dim": 128,
|
| 88 |
"linear_num_key_heads": 16,
|
|
|
|
| 110 |
"rope_theta": 10000000,
|
| 111 |
"partial_rotary_factor": 0.25
|
| 112 |
},
|
| 113 |
+
"bos_token_id": 248044,
|
| 114 |
+
"flm_version": "1.0.2"
|
| 115 |
+
}
|