config.json

{
  "_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": "\u0623\u062e\u0628\u0627\u0631 - \u062e\u0628\u0631",
    "1": "\u0625\u062c\u0627\u0628\u0629",
    "2": "\u0625\u062c\u0645\u0627\u0644\u064a",
    "3": "\u0625\u0633\u0627\u0621\u0629",
    "4": "\u0627\u0628\u062f\u0627\u0639 \u0641\u064a \u0627\u0644\u0639\u0645\u0644",
    "5": "\u0627\u062c\u062a\u0647\u0627\u062f",
    "6": "\u0627\u062d\u0635\u0644",
    "7": "\u0627\u0633\u0628\u0627\u0646\u064a\u0627",
    "8": "\u0627\u0633\u062a\u063a\u0631\u0627\u0642",
    "9": "\u0627\u0633\u0637\u0648\u0644 \u0628\u062d\u0631\u064a",
    "10": "\u0627\u0639\u062a\u0645\u0627\u062f \u0645\u0627\u0644\u064a",
    "11": "\u0627\u0642\u062a\u0646\u0627\u0621",
    "12": "\u0627\u0644\u062b\u0644\u0627\u062b\u0648\u0646",
    "13": "\u0627\u0644\u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u062d\u0633\u0627\u0628\u064a\u0629",
    "14": "\u0627\u0645\u064a\u0631 \u0627\u0644\u0628\u0644\u0627\u062f - \u0631\u0626\u064a\u0633 \u0627\u0644\u062c\u0645\u0647\u0648\u0631\u064a\u0629",
    "15": "\u062f\u0642\u064a\u0642\u0629",
    "16": "\u0630\u0647\u0628\u064a",
    "17": "\u0631\u062f\u0627\u0621",
    "18": "\u0635\u0627\u0626\u063a",
    "19": "\u0645\u0639\u064a\u0646"
  },
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "\u0623\u062e\u0628\u0627\u0631 - \u062e\u0628\u0631": 0,
    "\u0625\u062c\u0627\u0628\u0629": 1,
    "\u0625\u062c\u0645\u0627\u0644\u064a": 2,
    "\u0625\u0633\u0627\u0621\u0629": 3,
    "\u0627\u0628\u062f\u0627\u0639 \u0641\u064a \u0627\u0644\u0639\u0645\u0644": 4,
    "\u0627\u062c\u062a\u0647\u0627\u062f": 5,
    "\u0627\u062d\u0635\u0644": 6,
    "\u0627\u0633\u0628\u0627\u0646\u064a\u0627": 7,
    "\u0627\u0633\u062a\u063a\u0631\u0627\u0642": 8,
    "\u0627\u0633\u0637\u0648\u0644 \u0628\u062d\u0631\u064a": 9,
    "\u0627\u0639\u062a\u0645\u0627\u062f \u0645\u0627\u0644\u064a": 10,
    "\u0627\u0642\u062a\u0646\u0627\u0621": 11,
    "\u0627\u0644\u062b\u0644\u0627\u062b\u0648\u0646": 12,
    "\u0627\u0644\u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u062d\u0633\u0627\u0628\u064a\u0629": 13,
    "\u0627\u0645\u064a\u0631 \u0627\u0644\u0628\u0644\u0627\u062f - \u0631\u0626\u064a\u0633 \u0627\u0644\u062c\u0645\u0647\u0648\u0631\u064a\u0629": 14,
    "\u062f\u0642\u064a\u0642\u0629": 15,
    "\u0630\u0647\u0628\u064a": 16,
    "\u0631\u062f\u0627\u0621": 17,
    "\u0635\u0627\u0626\u063a": 18,
    "\u0645\u0639\u064a\u0646": 19
  },
  "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": "4.33.0",
  "tubelet_size": 2,
  "use_mean_pooling": false
}