{
  "architectures": [
    "VitDetBackbone"
  ],
  "attention_probs_dropout_prob": 0.1,
  "drop_path_rate": 0.0,
  "dropout_prob": 0.0,
  "dtype": "float32",
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.1,
  "hidden_size": 32,
  "image_size": 30,
  "initializer_range": 0.02,
  "intermediate_size": 37,
  "is_decoder": false,
  "layer_norm_eps": 1e-06,
  "mlp_ratio": 4,
  "model_type": "vitdet",
  "num_attention_heads": 4,
  "num_channels": 3,
  "num_hidden_layers": 2,
  "out_features": [
    "stage2"
  ],
  "out_indices": [
    2
  ],
  "patch_size": 2,
  "pretrain_image_size": 30,
  "qkv_bias": true,
  "residual_block_indices": [],
  "stage_names": [
    "stem",
    "stage1",
    "stage2"
  ],
  "transformers_version": "5.16.0.dev0",
  "use_absolute_position_embeddings": true,
  "use_relative_position_embeddings": false,
  "window_block_indices": [],
  "window_size": 0
}