{ | |
"_name_or_path": "google/vivit-b-16x2-kinetics400", | |
"architectures": [ | |
"VivitForVideoClassification" | |
], | |
"attention_probs_dropout_prob": 0.2, | |
"hidden_act": "gelu_fast", | |
"hidden_dropout_prob": 0.2, | |
"hidden_size": 768, | |
"id2label": { | |
"0": "cover", | |
"1": "defence", | |
"2": "flick", | |
"3": "hook", | |
"4": "lofted", | |
"5": "pull" | |
}, | |
"image_size": 224, | |
"initializer_range": 0.1, | |
"intermediate_size": 3072, | |
"label2id": { | |
"cover": 0, | |
"defence": 1, | |
"flick": 2, | |
"hook": 3, | |
"lofted": 4, | |
"pull": 5 | |
}, | |
"layer_norm_eps": 1e-12, | |
"model_type": "vivit", | |
"num_attention_heads": 12, | |
"num_channels": 3, | |
"num_frames": 32, | |
"num_hidden_layers": 12, | |
"qkv_bias": true, | |
"torch_dtype": "float32", | |
"transformers_version": "4.47.0", | |
"tubelet_size": [ | |
2, | |
16, | |
16 | |
], | |
"video_size": [ | |
32, | |
224, | |
224 | |
] | |
} | |