| { | |
| "_commit_hash": null, | |
| "_name_or_path": "./tiny-random-tvp", | |
| "distance_loss_weight": 1.0, | |
| "architectures": [ | |
| "TvpForVideoGrounding" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "backbone_config": { | |
| "architectures": [ | |
| "ResNetModel" | |
| ], | |
| "depths": [ | |
| 2, | |
| 2 | |
| ], | |
| "downsample_in_bottleneck": true, | |
| "hidden_sizes": [ | |
| 64, | |
| 128 | |
| ], | |
| "model_type": "resnet", | |
| "out_features": [ | |
| "stage2" | |
| ], | |
| "out_indices": [ | |
| 2 | |
| ], | |
| "stage_names": [ | |
| "stem", | |
| "stage1", | |
| "stage2" | |
| ], | |
| "torch_dtype": "float32" | |
| }, | |
| "duration_loss_weight": 0.1, | |
| "classifier": "mlp", | |
| "classifier_dropout": null, | |
| "cls_hidden_scale": 2, | |
| "hidden_act": "gelu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 128, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 384, | |
| "layer_norm_eps": 1e-12, | |
| "max_grid_col_position_embeddings": 100, | |
| "max_grid_row_position_embeddings": 100, | |
| "max_img_size": 448, | |
| "max_position_embeddings": 128, | |
| "max_text_length": 20, | |
| "model_type": "tvp", | |
| "num_attention_heads": 4, | |
| "num_frames": 1, | |
| "num_hidden_layers": 4, | |
| "visual_prompt_size": 96, | |
| "pad_token_id": 0, | |
| "position_embedding_type": "absolute", | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.29.2", | |
| "type_vocab_size": 2, | |
| "use_cache": true, | |
| "visual_prompter_apply": "replace", | |
| "visual_prompter_type": "framepad", | |
| "vocab_size": 30522, | |
| "vp_apply": "replace", | |
| "vp_type": "framepad" | |
| } | |