checkpoint/config.json

{
  "_name_or_path": "MCG-NJU/videomae-base-finetuned-kinetics",
  "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": "aam",
    "1": "aaple",
    "2": "ac",
    "3": "aids",
    "4": "alu",
    "5": "anaros",
    "6": "angur",
    "7": "apartment",
    "8": "attio",
    "9": "audio cassette",
    "10": "ayna",
    "11": "baandej",
    "12": "baat",
    "13": "baba",
    "14": "balti",
    "15": "balu",
    "16": "bhai",
    "17": "biscuts",
    "18": "bon",
    "19": "boroi",
    "20": "bottam",
    "21": "bou",
    "22": "cake",
    "23": "capsule",
    "24": "cha",
    "25": "chacha",
    "26": "chachi",
    "27": "chadar",
    "28": "chal",
    "29": "chikissha",
    "30": "chini",
    "31": "chips",
    "32": "chiruni",
    "33": "chocolate",
    "34": "chokh utha",
    "35": "chosma",
    "36": "churi",
    "37": "clip",
    "38": "cream",
    "39": "dada",
    "40": "dadi",
    "41": "daeitto",
    "42": "dal",
    "43": "debor",
    "44": "denadar",
    "45": "dengue",
    "46": "doctor",
    "47": "dongson",
    "48": "dulavai",
    "49": "durbol",
    "50": "jomoj",
    "51": "juta",
    "52": "konna",
    "53": "maa",
    "54": "tattha",
    "55": "toothpaste",
    "56": "tshirt",
    "57": "tubelight",
    "58": "tupi",
    "59": "tv"
  },
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "aam": 0,
    "aaple": 1,
    "ac": 2,
    "aids": 3,
    "alu": 4,
    "anaros": 5,
    "angur": 6,
    "apartment": 7,
    "attio": 8,
    "audio cassette": 9,
    "ayna": 10,
    "baandej": 11,
    "baat": 12,
    "baba": 13,
    "balti": 14,
    "balu": 15,
    "bhai": 16,
    "biscuts": 17,
    "bon": 18,
    "boroi": 19,
    "bottam": 20,
    "bou": 21,
    "cake": 22,
    "capsule": 23,
    "cha": 24,
    "chacha": 25,
    "chachi": 26,
    "chadar": 27,
    "chal": 28,
    "chikissha": 29,
    "chini": 30,
    "chips": 31,
    "chiruni": 32,
    "chocolate": 33,
    "chokh utha": 34,
    "chosma": 35,
    "churi": 36,
    "clip": 37,
    "cream": 38,
    "dada": 39,
    "dadi": 40,
    "daeitto": 41,
    "dal": 42,
    "debor": 43,
    "denadar": 44,
    "dengue": 45,
    "doctor": 46,
    "dongson": 47,
    "dulavai": 48,
    "durbol": 49,
    "jomoj": 50,
    "juta": 51,
    "konna": 52,
    "maa": 53,
    "tattha": 54,
    "toothpaste": 55,
    "tshirt": 56,
    "tubelight": 57,
    "tupi": 58,
    "tv": 59
  },
  "layer_norm_eps": 1e-12,
  "model_type": "videomae",
  "norm_pix_loss": false,
  "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": true
}