{
  "_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": "help",
    "1": "like",
    "2": "later",
    "3": "bird",
    "4": "bowling",
    "5": "walk",
    "6": "same",
    "7": "work",
    "8": "can",
    "9": "tall",
    "10": "language",
    "11": "enjoy",
    "12": "color",
    "13": "cook",
    "14": "drink",
    "15": "purple",
    "16": "black",
    "17": "but",
    "18": "many",
    "19": "fine",
    "20": "secretary",
    "21": "deaf",
    "22": "year",
    "23": "orange",
    "24": "candy",
    "25": "dark",
    "26": "bed",
    "27": "pizza",
    "28": "paint",
    "29": "book",
    "30": "hat",
    "31": "cow",
    "32": "cheat",
    "33": "all",
    "34": "man",
    "35": "give",
    "36": "cool",
    "37": "table",
    "38": "right",
    "39": "forget",
    "40": "apple",
    "41": "letter",
    "42": "hearing",
    "43": "go",
    "44": "full",
    "45": "doctor",
    "46": "dog",
    "47": "family",
    "48": "yes",
    "49": "hot",
    "50": "brown",
    "51": "corn",
    "52": "how",
    "53": "cousin",
    "54": "basketball",
    "55": "son",
    "56": "africa",
    "57": "thursday",
    "58": "pink",
    "59": "change",
    "60": "meet",
    "61": "computer",
    "62": "white",
    "63": "play",
    "64": "chair",
    "65": "tell",
    "66": "need",
    "67": "finish",
    "68": "woman",
    "69": "want",
    "70": "decide",
    "71": "kiss",
    "72": "paper",
    "73": "jacket",
    "74": "eat",
    "75": "last",
    "76": "time",
    "77": "fish",
    "78": "graduate",
    "79": "birthday",
    "80": "thin",
    "81": "what",
    "82": "city",
    "83": "no",
    "84": "short",
    "85": "clothes",
    "86": "pull",
    "87": "dance",
    "88": "shirt",
    "89": "medicine",
    "90": "mother",
    "91": "before",
    "92": "now",
    "93": "wrong",
    "94": "accident",
    "95": "who",
    "96": "blue",
    "97": "study",
    "98": "school",
    "99": "thanksgiving"
  },
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "accident": 94,
    "africa": 56,
    "all": 33,
    "apple": 40,
    "basketball": 54,
    "bed": 26,
    "before": 91,
    "bird": 3,
    "birthday": 79,
    "black": 16,
    "blue": 96,
    "book": 29,
    "bowling": 4,
    "brown": 50,
    "but": 17,
    "can": 8,
    "candy": 24,
    "chair": 64,
    "change": 59,
    "cheat": 32,
    "city": 82,
    "clothes": 85,
    "color": 12,
    "computer": 61,
    "cook": 13,
    "cool": 36,
    "corn": 51,
    "cousin": 53,
    "cow": 31,
    "dance": 87,
    "dark": 25,
    "deaf": 21,
    "decide": 70,
    "doctor": 45,
    "dog": 46,
    "drink": 14,
    "eat": 74,
    "enjoy": 11,
    "family": 47,
    "fine": 19,
    "finish": 67,
    "fish": 77,
    "forget": 39,
    "full": 44,
    "give": 35,
    "go": 43,
    "graduate": 78,
    "hat": 30,
    "hearing": 42,
    "help": 0,
    "hot": 49,
    "how": 52,
    "jacket": 73,
    "kiss": 71,
    "language": 10,
    "last": 75,
    "later": 2,
    "letter": 41,
    "like": 1,
    "man": 34,
    "many": 18,
    "medicine": 89,
    "meet": 60,
    "mother": 90,
    "need": 66,
    "no": 83,
    "now": 92,
    "orange": 23,
    "paint": 28,
    "paper": 72,
    "pink": 58,
    "pizza": 27,
    "play": 63,
    "pull": 86,
    "purple": 15,
    "right": 38,
    "same": 6,
    "school": 98,
    "secretary": 20,
    "shirt": 88,
    "short": 84,
    "son": 55,
    "study": 97,
    "table": 37,
    "tall": 9,
    "tell": 65,
    "thanksgiving": 99,
    "thin": 80,
    "thursday": 57,
    "time": 76,
    "walk": 5,
    "want": 69,
    "what": 81,
    "white": 62,
    "who": 95,
    "woman": 68,
    "work": 7,
    "wrong": 93,
    "year": 22,
    "yes": 48
  },
  "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": "4.46.1",
  "tubelet_size": 2,
  "use_mean_pooling": false
}