MBZUAI/GLaMM-FullScope_v0

1{2  "_name_or_path": "./trained_checkpoints/exp1_grand_pretrain_v10_gs4488_ckpt_25000_hf",3  "architectures": [4    "LISAForCausalLM"5  ],6  "bbox_token_idx": 32002,7  "bos_token_id": 1,8  "eos_token_id": 2,9  "freeze_mm_mlp_adapter": true,10  "hidden_act": "silu",11  "hidden_size": 4096,12  "image_aspect_ratio": "square",13  "image_grid_pinpoints": null,14  "initializer_range": 0.02,15  "intermediate_size": 11008,16  "max_position_embeddings": 2048,17  "mm_hidden_size": 1024,18  "mm_use_im_patch_token": false,19  "mm_use_im_start_end": true,20  "mm_vision_select_feature": "patch",21  "mm_vision_select_layer": -2,22  "mm_vision_tower": "openai/clip-vit-large-patch14",23  "model_type": "llava",24  "num_attention_heads": 32,25  "num_hidden_layers": 32,26  "num_level_reg_features": 4,27  "out_dim": 256,28  "pad_token_id": 0,29  "pretrain_mm_mlp_adapter": null,30  "rms_norm_eps": 1e-06,31  "sep_image_conv_front": false,32  "tie_word_embeddings": false,33  "torch_dtype": "bfloat16",34  "train_mask_decoder": true,35  "transformers_version": "4.28.0.dev0",36  "tune_mm_mlp_adapter": false,37  "use_cache": false,38  "use_mm_proj": true,39  "vision_tower": "openai/clip-vit-large-patch14",40  "vocab_size": 32009,41  "with_spi": true42}43