{
  "architectures": [
    "HieraForPreTraining"
  ],
  "decoder_depth": 8,
  "decoder_hidden_size": 512,
  "decoder_num_heads": 16,
  "depths": [
    2,
    6,
    36,
    4
  ],
  "drop_path_rate": 0.0,
  "embed_dim": 256,
  "embed_dim_multiplier": 2.0,
  "hidden_act": "gelu",
  "hidden_size": 2048,
  "image_size": [
    224,
    224
  ],
  "initializer_range": 0.02,
  "layer_norm_eps": 1e-06,
  "layer_norm_init": 1.0,
  "mask_ratio": 0.6,
  "masked_unit_attention": [
    true,
    true,
    false,
    false
  ],
  "masked_unit_size": [
    8,
    8
  ],
  "mlp_ratio": 4.0,
  "model_type": "hiera",
  "norm_pix_loss": true,
  "num_channels": 3,
  "num_heads": [
    4,
    8,
    16,
    32
  ],
  "num_layers": 4,
  "num_query_pool": 2,
  "out_features": [
    "stage4"
  ],
  "out_indices": [
    4
  ],
  "patch_padding": [
    3,
    3
  ],
  "patch_size": [
    7,
    7
  ],
  "patch_stride": [
    4,
    4
  ],
  "query_stride": [
    2,
    2
  ],
  "stage_names": [
    "stem",
    "stage1",
    "stage2",
    "stage3",
    "stage4"
  ],
  "torch_dtype": "float32",
  "transformers_version": "4.41.0.dev0",
  "use_separate_position_embedding": false
}