config.json
| 1 | { |
| 2 | "architectures": [ |
| 3 | "LlavaForConditionalGeneration" |
| 4 | ], |
| 5 | "ignore_index": -100, |
| 6 | "image_token_index": 32000, |
| 7 | "model_type": "llava", |
| 8 | "pad_token_id": 32001, |
| 9 | "projector_hidden_act": "gelu", |
| 10 | "text_config": { |
| 11 | "_name_or_path": "lmsys/vicuna-7b-v1.5", |
| 12 | "architectures": [ |
| 13 | "LlamaForCausalLM" |
| 14 | ], |
| 15 | "max_position_embeddings": 4096, |
| 16 | "model_type": "llama", |
| 17 | "rms_norm_eps": 1e-05, |
| 18 | "torch_dtype": "float16", |
| 19 | "vocab_size": 32064 |
| 20 | }, |
| 21 | "tie_word_embeddings": false, |
| 22 | "torch_dtype": "float16", |
| 23 | "transformers_version": "4.36.0.dev0", |
| 24 | "vision_config": { |
| 25 | "hidden_size": 1024, |
| 26 | "image_size": 336, |
| 27 | "intermediate_size": 4096, |
| 28 | "model_type": "clip_vision_model", |
| 29 | "num_attention_heads": 16, |
| 30 | "num_hidden_layers": 24, |
| 31 | "patch_size": 14, |
| 32 | "projection_dim": 768, |
| 33 | "vocab_size": 32000 |
| 34 | }, |
| 35 | "vision_feature_layer": -2, |
| 36 | "vision_feature_select_strategy": "default", |
| 37 | "vocab_size": 32064 |
| 38 | } |
| 39 | |