{
  "architectures": [
    "LevitForImageClassification"
  ],
  "attention_ratio": [
    2,
    2,
    2
  ],
  "depths": [
    2,
    3,
    4
  ],
  "down_ops": [
    [
      "Subsample",
      16,
      8,
      4,
      2,
      2
    ],
    [
      "Subsample",
      16,
      16,
      4,
      2,
      2
    ],
    [
      ""
    ]
  ],
  "drop_path_rate": 0,
  "hidden_sizes": [
    128,
    256,
    384
  ],
  "image_size": 64,
  "initializer_range": 0.02,
  "kernel_size": 3,
  "key_dim": [
    16,
    16,
    16
  ],
  "mlp_ratio": [
    2,
    2,
    2
  ],
  "model_type": "levit",
  "num_attention_heads": [
    4,
    6,
    8
  ],
  "num_channels": 3,
  "padding": 1,
  "patch_size": 16,
  "stride": 2,
  "torch_dtype": "float32",
  "transformers_version": "4.28.0.dev0"
}