Upload config.json with huggingface_hub
Browse files- config.json +45 -0
config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": ["MobilintQwen2VLForConditionalGeneration"],
|
| 3 |
+
"model_type": "mobilint-qwen2_vl",
|
| 4 |
+
"bos_token_id": 151643,
|
| 5 |
+
"eos_token_id": 151645,
|
| 6 |
+
"hidden_size": 1536,
|
| 7 |
+
"intermediate_size": 8960,
|
| 8 |
+
"max_position_embeddings": 4096,
|
| 9 |
+
"num_attention_heads": 12,
|
| 10 |
+
"num_hidden_layers": 28,
|
| 11 |
+
"num_key_value_heads": 2,
|
| 12 |
+
"rope_theta": 1000000.0,
|
| 13 |
+
"vocab_size": 151936,
|
| 14 |
+
"tie_word_embeddings": false,
|
| 15 |
+
"mxq_path": "Qwen2-VL-2B-Instruct_text_model.mxq",
|
| 16 |
+
"gguf_path": "qwen2-vl-2b-instruct-vocab.gguf",
|
| 17 |
+
"embd_path": "target_emb.bin",
|
| 18 |
+
"core_mode": "single",
|
| 19 |
+
"dev_no": 0,
|
| 20 |
+
"target_cores": ["0:0"],
|
| 21 |
+
"vision_start_token_id": 151652,
|
| 22 |
+
"vision_end_token_id": 151653,
|
| 23 |
+
"vision_token_id": 151654,
|
| 24 |
+
"image_token_id": 151655,
|
| 25 |
+
"video_token_id": 151656,
|
| 26 |
+
"vision_config": {
|
| 27 |
+
"depth": 32,
|
| 28 |
+
"embed_dim": 1280,
|
| 29 |
+
"mlp_ratio": 4,
|
| 30 |
+
"mxq_path": "Qwen2-VL-2B-Instruct_vision_transformer.mxq",
|
| 31 |
+
"num_heads": 16,
|
| 32 |
+
"in_chans": 3,
|
| 33 |
+
"hidden_size": 1536,
|
| 34 |
+
"patch_size": 14,
|
| 35 |
+
"spatial_merge_size": 2,
|
| 36 |
+
"spatial_patch_size": 14,
|
| 37 |
+
"temporal_patch_size": 2,
|
| 38 |
+
"core_mode": "multi",
|
| 39 |
+
"target_clusters": [1]
|
| 40 |
+
},
|
| 41 |
+
"rope_scaling": {
|
| 42 |
+
"type": "mrope",
|
| 43 |
+
"mrope_section": [16, 24, 24]
|
| 44 |
+
}
|
| 45 |
+
}
|