SurplusDeficit/GeLM
0
1{2 "_name_or_path": "./checkpoints/vicuna-v1-3-7b",3 "architectures": [4 "GelmLlamaForCausalLM"5 ],6 "attention_bias": false,7 "bos_token_id": 1,8 "eos_token_id": 2,9 "freeze_mm_mlp_adapter": false,10 "grounding_args": {11 "d_gnd": 1024,12 "d_proj": 1024,13 "gnd_enc_layers": 114 },15 "hidden_act": "silu",16 "hidden_size": 4096,17 "image_aspect_ratio": "square",18 "image_grid_pinpoints": null,19 "initializer_range": 0.02,20 "input_type": "feature",21 "intermediate_size": 11008,22 "max_position_embeddings": 2048,23 "mm_hidden_size": 768,24 "mm_use_im_patch_token": false,25 "mm_use_im_start_end": false,26 "mm_vision_select_feature": "patch",27 "mm_vision_select_layer": -2,28 "mm_vision_tower": "/remote-home/share/huggingface/clip-vit-large-patch14/",29 "model_type": "gelm",30 "num_attention_heads": 32,31 "num_frames": 100,32 "num_hidden_layers": 32,33 "num_key_value_heads": 32,34 "pad_token_id": 0,35 "pretraining_tp": 1,36 "rms_norm_eps": 1e-06,37 "rope_scaling": null,38 "rope_theta": 10000.0,39 "saliency_token_ids": [40 32010,41 3201142 ],43 "special_token_ids": [44 32000,45 32001,46 32002,47 32003,48 32004,49 32005,50 32006,51 32007,52 32008,53 3200954 ],55 "tie_word_embeddings": false,56 "torch_dtype": "bfloat16",57 "transformers_version": "4.34.0",58 "tune_mm_mlp_adapter": false,59 "use_cache": true,60 "use_mm_proj": true,61 "vocab_size": 3201262}63 