| { | |
| "_name_or_path": "/data/user/qxiao183/qxiao183test2/PLM-V/MM-Pretrain/models/LLaVA-PLM-1.8B-Instruct-hf", | |
| "architectures": [ | |
| "LlavaPLMForConditionalGeneration" | |
| ], | |
| "model_type": "llava_plm", | |
| "ignore_index": -100, | |
| "image_token_index": 151646, | |
| "projector_hidden_act": "gelu", | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.43.4", | |
| "vision_feature_layer": -2, | |
| "vision_feature_select_strategy": "full", | |
| "text_config": { | |
| "model_type": "plm", | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 151643, | |
| "eos_token_id": 151643, | |
| "hidden_act": "relu2", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 8192, | |
| "max_position_embeddings": 4096, | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 32, | |
| "num_key_value_heads": 16, | |
| "pretraining_tp": 1, | |
| "qk_nope_head_dim": 128, | |
| "qk_rope_head_dim": 64, | |
| "rms_norm_eps": 1e-06, | |
| "rope_scaling": null, | |
| "rope_theta": 100000.0, | |
| "use_cache": true, | |
| "v_head_dim": 128, | |
| "vocab_size": 151936 | |
| }, | |
| "vision_config": { | |
| "model_type": "siglip_vision_model", | |
| "hidden_size": 1152, | |
| "image_size": 384, | |
| "intermediate_size": 4304, | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 27, | |
| "patch_size": 16 | |
| } | |
| } |