OpenGVLab/VideoChat-Flash-Qwen2-7B_res448

1{2  "architectures": [3    "VideoChatFlashQwenForCausalLM"4  ],5  "auto_map": {6    "AutoConfig": "modeling_videochat_flash.VideoChatFlashQwenConfig",7    "AutoModel": "modeling_videochat_flash.VideoChatFlashQwenForCausalLM"8  },9  "attention_dropout": 0.0,10  "bos_token_id": 151643,11  "eos_token_id": 151645,12  "frame_aspect_ratio": "square",13  "frame_grid_pinpoints": null,14  "hidden_act": "silu",15  "hidden_size": 3584,16  "image_aspect_ratio": "anyres_nopad",17  "image_crop_resolution": null,18  "image_grid_pinpoints": [19    [20      448,21      44822    ],23    [24      448,25      89626    ],27    [28      448,29      134430    ],31    [32      448,33      179234    ],35    [36      448,37      224038    ],39    [40      448,41      268842    ],43    [44      896,45      44846    ],47    [48      896,49      89650    ],51    [52      896,53      134454    ],55    [56      896,57      179258    ],59    [60      896,61      224062    ],63    [64      896,65      268866    ],67    [68      1344,69      44870    ],71    [72      1344,73      89674    ],75    [76      1344,77      134478    ],79    [80      1344,81      179282    ],83    [84      1344,85      224086    ],87    [88      1344,89      268890    ],91    [92      1792,93      44894    ],95    [96      1792,97      89698    ],99    [100      1792,101      1344102    ],103    [104      1792,105      1792106    ],107    [108      1792,109      2240110    ],111    [112      1792,113      2688114    ],115    [116      2240,117      448118    ],119    [120      2240,121      896122    ],123    [124      2240,125      1344126    ],127    [128      2240,129      1792130    ],131    [132      2240,133      2240134    ],135    [136      2240,137      2688138    ],139    [140      2688,141      448142    ],143    [144      2688,145      896146    ],147    [148      2688,149      1344150    ],151    [152      2688,153      1792154    ],155    [156      2688,157      2240158    ],159    [160      2688,161      2688162    ]163  ],164  "image_split_resolution": null,165  "initializer_range": 0.02,166  "intermediate_size": 18944,167  "llm_compress_layer_list": [168    24169  ],170  "llm_compress_type": "attention",171  "llm_image_token_ratio_list": [172    1.0,173 174    0.5175  ],176  "max_num_pixels": 14745600000,177  "max_position_embeddings": 32768,178  "max_window_layers": 28,179  "min_slow_num_frames": 4,180  "mm_close_init": false,181  "mm_hidden_size": 1024,182  "mm_llm_compress": false,183  "mm_local_num_frames": 4,184  "mm_newline_position": "nothing",185  "mm_num_compress_latents": 128,186  "mm_num_compress_query_type": "learnable",187  "mm_patch_merge_type": "spatial_nopad",188  "mm_pos_num_frames": 8,189  "mm_projector_lr": null,190  "mm_projector_type": "tome16_mlp_hd64",191  "mm_resampler_type": null,192  "mm_spatial_pool_mode": "bilinear",193  "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter",194  "mm_use_im_patch_token": false,195  "mm_use_im_start_end": false,196  "mm_vision_select_feature": "patch",197  "mm_vision_select_layer": -2,198  "mm_vision_tower": "umt-hd-large",199  "mm_vision_tower_lr": 2e-06,200  "model_type": "videochat_flash_qwen",201  "num_attention_heads": 28,202  "num_hidden_layers": 28,203  "num_key_value_heads": 4,204  "pos_skipping_range": 4096,205  "rms_norm_eps": 1e-06,206  "rope_scaling": null,207  "rope_theta": 1000000.0,208  "sliding_window": 131072,209  "tie_word_embeddings": false,210  "tokenizer_model_max_length": 32768,211  "tokenizer_padding_side": "right",212  "torch_dtype": "bfloat16",213  "transformers_version": "4.40.1",214  "use_cache": true,215  "use_mm_proj": true,216  "use_pos_skipping": false,217  "use_sliding_window": false,218  "vision_encode_type": "video_image",219  "vision_tower_pretrained": null,220  "vocab_size": 152064221}222