|
{ |
|
"_name_or_path": "MCG-NJU/videomae-base", |
|
"architectures": [ |
|
"VideoMAEForVideoClassification" |
|
], |
|
"attention_probs_dropout_prob": 0.0, |
|
"decoder_hidden_size": 384, |
|
"decoder_intermediate_size": 1536, |
|
"decoder_num_attention_heads": 6, |
|
"decoder_num_hidden_layers": 4, |
|
"hidden_act": "gelu", |
|
"hidden_dropout_prob": 0.0, |
|
"hidden_size": 768, |
|
"id2label": { |
|
"0": "1_1", |
|
"1": "1_2", |
|
"2": "1_3", |
|
"3": "1_4", |
|
"4": "1_5", |
|
"5": "1_6", |
|
"6": "1_7", |
|
"7": "1_8", |
|
"8": "1_9", |
|
"9": "1_nan", |
|
"10": "5_1", |
|
"11": "5_2", |
|
"12": "5_3", |
|
"13": "5_4", |
|
"14": "5_5", |
|
"15": "5_6", |
|
"16": "5_7", |
|
"17": "5_8", |
|
"18": "5_9", |
|
"19": "5_nan", |
|
"20": "6_1", |
|
"21": "6_2", |
|
"22": "6_3", |
|
"23": "6_4", |
|
"24": "6_5", |
|
"25": "6_6", |
|
"26": "6_7", |
|
"27": "6_8", |
|
"28": "6_9", |
|
"29": "7_1", |
|
"30": "7_2", |
|
"31": "7_3", |
|
"32": "7_5", |
|
"33": "7_6", |
|
"34": "7_7", |
|
"35": "7_8", |
|
"36": "7_9", |
|
"37": "9_1", |
|
"38": "9_3", |
|
"39": "9_4", |
|
"40": "9_5", |
|
"41": "9_6", |
|
"42": "9_7", |
|
"43": "9_8", |
|
"44": "9_9", |
|
"45": "nan_nan" |
|
}, |
|
"image_size": 224, |
|
"initializer_range": 0.02, |
|
"intermediate_size": 3072, |
|
"label2id": { |
|
"1_1": 0, |
|
"1_2": 1, |
|
"1_3": 2, |
|
"1_4": 3, |
|
"1_5": 4, |
|
"1_6": 5, |
|
"1_7": 6, |
|
"1_8": 7, |
|
"1_9": 8, |
|
"1_nan": 9, |
|
"5_1": 10, |
|
"5_2": 11, |
|
"5_3": 12, |
|
"5_4": 13, |
|
"5_5": 14, |
|
"5_6": 15, |
|
"5_7": 16, |
|
"5_8": 17, |
|
"5_9": 18, |
|
"5_nan": 19, |
|
"6_1": 20, |
|
"6_2": 21, |
|
"6_3": 22, |
|
"6_4": 23, |
|
"6_5": 24, |
|
"6_6": 25, |
|
"6_7": 26, |
|
"6_8": 27, |
|
"6_9": 28, |
|
"7_1": 29, |
|
"7_2": 30, |
|
"7_3": 31, |
|
"7_5": 32, |
|
"7_6": 33, |
|
"7_7": 34, |
|
"7_8": 35, |
|
"7_9": 36, |
|
"9_1": 37, |
|
"9_3": 38, |
|
"9_4": 39, |
|
"9_5": 40, |
|
"9_6": 41, |
|
"9_7": 42, |
|
"9_8": 43, |
|
"9_9": 44, |
|
"nan_nan": 45 |
|
}, |
|
"layer_norm_eps": 1e-12, |
|
"model_type": "videomae", |
|
"norm_pix_loss": true, |
|
"num_attention_heads": 12, |
|
"num_channels": 3, |
|
"num_frames": 16, |
|
"num_hidden_layers": 12, |
|
"patch_size": 16, |
|
"problem_type": "single_label_classification", |
|
"qkv_bias": true, |
|
"torch_dtype": "float32", |
|
"transformers_version": "4.44.0", |
|
"tubelet_size": 2, |
|
"use_mean_pooling": false |
|
} |
|
|