{
  "_name_or_path": "MCG-NJU/videomae-base-short-ssv2",
  "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": "\u0627\u0644\u0627\u0633\u0641\u0644_0.mp4",
    "1": "\u0627\u0644\u0627\u0639\u0644\u0649_0.mp4",
    "2": "\u0627\u0644\u0627\u0645\u0627\u0645_0.mp4",
    "3": "\u0627\u0644\u062e\u0644\u0641_0.mp4",
    "4": "\u0627\u0644\u064a\u0633\u0627\u0631_0.mp4",
    "5": "\u0627\u0644\u064a\u0645\u064a\u0646_0.mp4",
    "6": "\u062a\u062d\u062a_0.mp4",
    "7": "\u062c\u0627\u0646\u0628_0.mp4",
    "8": "\u062c\u0646\u0648\u0628_0.mp4",
    "9": "\u062d\u0648\u0644_0.mp4",
    "10": "\u0634\u0631\u0642_0.mp4",
    "11": "\u0634\u0645\u0627\u0644_0.mp4",
    "12": "\u0639\u0646\u062f_0.mp4",
    "13": "\u063a\u0631\u0628_0.mp4",
    "14": "\u0641\u0648\u0642_0.mp4",
    "15": "\u0642\u0645\u0629_0.mp4",
    "16": "\u0645\u0642\u0627\u0628\u0644_0.mp4"
  },
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "\u0627\u0644\u0627\u0633\u0641\u0644_0.mp4": 0,
    "\u0627\u0644\u0627\u0639\u0644\u0649_0.mp4": 1,
    "\u0627\u0644\u0627\u0645\u0627\u0645_0.mp4": 2,
    "\u0627\u0644\u062e\u0644\u0641_0.mp4": 3,
    "\u0627\u0644\u064a\u0633\u0627\u0631_0.mp4": 4,
    "\u0627\u0644\u064a\u0645\u064a\u0646_0.mp4": 5,
    "\u062a\u062d\u062a_0.mp4": 6,
    "\u062c\u0627\u0646\u0628_0.mp4": 7,
    "\u062c\u0646\u0648\u0628_0.mp4": 8,
    "\u062d\u0648\u0644_0.mp4": 9,
    "\u0634\u0631\u0642_0.mp4": 10,
    "\u0634\u0645\u0627\u0644_0.mp4": 11,
    "\u0639\u0646\u062f_0.mp4": 12,
    "\u063a\u0631\u0628_0.mp4": 13,
    "\u0641\u0648\u0642_0.mp4": 14,
    "\u0642\u0645\u0629_0.mp4": 15,
    "\u0645\u0642\u0627\u0628\u0644_0.mp4": 16
  },
  "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": "4.33.2",
  "tubelet_size": 2,
  "use_mean_pooling": false
}