{ "_name_or_path": "MCG-NJU/videomae-large", "architectures": [ "VideoMAEForVideoClassification" ], "attention_probs_dropout_prob": 0.0, "decoder_hidden_size": 512, "decoder_intermediate_size": 2048, "decoder_num_attention_heads": 8, "decoder_num_hidden_layers": 12, "hidden_act": "gelu", "hidden_dropout_prob": 0.0, "hidden_size": 1024, "id2label": { "0": "basketball", "1": "biking", "2": "diving", "3": "golf_swing", "4": "horse_riding", "5": "soccer_juggling", "6": "swing", "7": "tennis_swing", "8": "trampoline_jumping", "9": "volleyball_spiking", "10": "walking" }, "image_size": 224, "initializer_range": 0.02, "intermediate_size": 4096, "label2id": { "basketball": 0, "biking": 1, "diving": 2, "golf_swing": 3, "horse_riding": 4, "soccer_juggling": 5, "swing": 6, "tennis_swing": 7, "trampoline_jumping": 8, "volleyball_spiking": 9, "walking": 10 }, "layer_norm_eps": 1e-12, "model_type": "videomae", "norm_pix_loss": true, "num_attention_heads": 16, "num_channels": 3, "num_frames": 16, "num_hidden_layers": 24, "patch_size": 16, "problem_type": "single_label_classification", "qkv_bias": true, "torch_dtype": "float32", "transformers_version": "4.39.3", "tubelet_size": 2, "use_mean_pooling": false }