| { | |
| "architectures": ["LlavaLlamaModel"], | |
| "model_type": "llava_llama", | |
| "chat_template": null, | |
| "torch_dtype": "bfloat16", | |
| "hidden_size": 4096, | |
| "mm_hidden_size": 1152, | |
| "num_video_frames": 8, | |
| "image_aspect_ratio": "resize", | |
| "vision_resolution": -1, | |
| "mm_vision_select_feature": "cls_patch", | |
| "mm_vision_select_layer": -2, | |
| "mm_projector_type": "mlp_downsample", | |
| "mm_use_im_patch_token": false, | |
| "mm_use_im_start_end": false, | |
| "model_dtype": "torch.bfloat16", | |
| "num_time_tokens": 0, | |
| "time_token_format": "<t{t}>", | |
| "time_token_ids": [], | |
| "tune_language_model": true, | |
| "tune_mm_projector": true, | |
| "tune_vision_tower": false, | |
| "mm_projector_cfg": { | |
| "model_type": "v2l_projector", | |
| "architectures": ["MultimodalProjector"], | |
| "mm_projector_type": "mlp_downsample", | |
| "hidden_size": 1152, | |
| "output_attentions": false, | |
| "output_hidden_states": false, | |
| "torch_dtype": "bfloat16" | |
| }, | |
| "vision_tower_cfg": { | |
| "model_type": "siglip_vision_model", | |
| "architectures": ["SiglipVisionModel"], | |
| "image_size": 384, | |
| "patch_size": 14, | |
| "num_hidden_layers": 27, | |
| "hidden_size": 1152, | |
| "num_attention_heads": 16, | |
| "intermediate_size": 4304, | |
| "layer_norm_eps": 1e-06, | |
| "num_channels": 3, | |
| "torch_dtype": "bfloat16" | |
| }, | |
| "llm_config": { | |
| "model_type": "llama", | |
| "architectures": ["LlamaForCausalLM"], | |
| "vocab_size": 128259, | |
| "hidden_size": 4096, | |
| "num_hidden_layers": 32, | |
| "num_attention_heads": 32, | |
| "num_key_value_heads": 8, | |
| "intermediate_size": 14336, | |
| "rms_norm_eps": 1e-05, | |
| "max_position_embeddings": 8192, | |
| "model_max_length": 4096, | |
| "bos_token_id": 128000, | |
| "eos_token_id": 128009, | |
| "pad_token_id": null, | |
| "rope_theta": 500000.0, | |
| "tie_word_embeddings": false, | |
| "use_cache": true, | |
| "torch_dtype": "bfloat16" | |
| } | |
| } | |