MLAdaptiveIntelligence/LLaVAction-7B
196
1{2 "_name_or_path": "lmms-lab/LLaVA-Video-7B-Qwen2",3 "action_types": "97,300,3806",4 "add_faster_video": false,5 "add_time_instruction": true,6 "architectures": [7 "LlavaQwenForCausalLM"8 ],9 "attention_dropout": 0.0,10 "bos_token_id": 151643,11 "eos_token_id": 151645,12 "faster_token_stride": 10,13 "force_sample": false,14 "hidden_act": "silu",15 "hidden_size": 3584,16 "image_aspect_ratio": "anyres_max_9",17 "image_crop_resolution": null,18 "image_grid_pinpoints": [19 [20 384,21 38422 ],23 [24 384,25 76826 ],27 [28 384,29 115230 ],31 [32 384,33 153634 ],35 [36 384,37 192038 ],39 [40 384,41 230442 ],43 [44 768,45 38446 ],47 [48 768,49 76850 ],51 [52 768,53 115254 ],55 [56 768,57 153658 ],59 [60 768,61 192062 ],63 [64 768,65 230466 ],67 [68 1152,69 38470 ],71 [72 1152,73 76874 ],75 [76 1152,77 115278 ],79 [80 1152,81 153682 ],83 [84 1152,85 192086 ],87 [88 1152,89 230490 ],91 [92 1536,93 38494 ],95 [96 1536,97 76898 ],99 [100 1536,101 1152102 ],103 [104 1536,105 1536106 ],107 [108 1536,109 1920110 ],111 [112 1536,113 2304114 ],115 [116 1920,117 384118 ],119 [120 1920,121 768122 ],123 [124 1920,125 1152126 ],127 [128 1920,129 1536130 ],131 [132 1920,133 1920134 ],135 [136 1920,137 2304138 ],139 [140 2304,141 384142 ],143 [144 2304,145 768146 ],147 [148 2304,149 1152150 ],151 [152 2304,153 1536154 ],155 [156 2304,157 1920158 ],159 [160 2304,161 2304162 ]163 ],164 "image_split_resolution": null,165 "image_token_index": 151646,166 "initializer_range": 0.02,167 "intermediate_size": 18944,168 "max_position_embeddings": 32768,169 "max_window_layers": 28,170 "mm_hidden_size": 1152,171 "mm_newline_position": "grid",172 "mm_patch_merge_type": "spatial_unpad",173 "mm_projector_lr": null,174 "mm_projector_type": "mlp2x_gelu",175 "mm_resampler_type": null,176 "mm_spatial_pool_mode": "bilinear",177 "mm_spatial_pool_stride": null,178 "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",179 "mm_use_im_patch_token": false,180 "mm_use_im_start_end": false,181 "mm_vision_select_feature": "patch",182 "mm_vision_select_layer": -2,183 "mm_vision_tower": "google/siglip-so400m-patch14-384",184 "mm_vision_tower_lr": 2e-06,185 "model_type": "qwen2",186 "num_attention_heads": 28,187 "num_hidden_layers": 28,188 "num_key_value_heads": 4,189 "pos_skipping_range": 4096,190 "rms_norm_eps": 1e-06,191 "rope_scaling": null,192 "rope_theta": 1000000.0,193 "sliding_window": 131072,194 "tie_word_embeddings": false,195 "tokenizer_model_max_length": 32768,196 "tokenizer_padding_side": "right",197 "torch_dtype": "bfloat16",198 "transformers_version": "4.40.0.dev0",199 "use_cache": true,200 "use_mm_proj": true,201 "use_pos_skipping": false,202 "use_sliding_window": false,203 "vision_supervision": "one_token",204 "vision_token_training": "last_layer",205 "vision_tower_pretrained": null,206 "vocab_size": 152064207}208 