config.json

{
  "architectures": [
    "VideoMAEForPreTraining"
  ],
  "attention_probs_dropout_prob": 0.0,
  "attn_implementation": "flash_attention_2",
  "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,
  "image_size": 384,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "layer_norm_eps": 1e-12,
  "model_type": "videomae",
  "norm_pix_loss": true,
  "num_attention_heads": 12,
  "num_channels": 1,
  "num_frames": 320,
  "num_hidden_layers": 12,
  "patch_size": 16,
  "qkv_bias": true,
  "torch_dtype": "float32",
  "transformers_version": "4.46.1",
  "tubelet_size": 16,
  "use_mean_pooling": true
}