config.json

{
  "apply_layernorm": true,
  "architectures": [
    "DepthProForDepthEstimation"
  ],
  "attention_probs_dropout_prob": 0.0,
  "drop_path_rate": 0.0,
  "fusion_hidden_size": 256,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 1024,
  "image_size": 1536,
  "initializer_range": 0.02,
  "intermediate_feature_dims": [
    256,
    256
  ],
  "intermediate_hook_ids": [
    11,
    5
  ],
  "layer_norm_eps": 1e-06,
  "layerscale_value": 1.0,
  "mlp_ratio": 4,
  "model_type": "depth_pro",
  "num_attention_heads": 16,
  "num_channels": 3,
  "num_fov_head_layers": 2,
  "num_hidden_layers": 24,
  "patch_embeddings_size": 16,
  "patch_size": 384,
  "qkv_bias": true,
  "reshape_hidden_states": true,
  "scaled_images_feature_dims": [
    1024,
    1024,
    512
  ],
  "scaled_images_overlap_ratios": [
    0.0,
    0.5,
    0.25
  ],
  "scaled_images_ratios": [
    0.25,
    0.5,
    1
  ],
  "torch_dtype": "float16",
  "transformers_version": "4.47.0.dev0",
  "use_batch_norm_in_fusion_residual": false,
  "use_bias_in_fusion_residual": true,
  "use_fov_model": true,
  "use_swiglu_ffn": false
}