{ "architectures": [ "DepthProForDepthEstimation" ], "fov_model_config": { "hidden_size": 1024, "image_size": 384, "model_type": "dinov2", "num_attention_heads": 16, "num_hidden_layers": 24, "out_features": [ "stage24" ], "out_indices": [ 24 ], "patch_size": 16, "stage_names": [ "stem", "stage1", "stage2", "stage3", "stage4", "stage5", "stage6", "stage7", "stage8", "stage9", "stage10", "stage11", "stage12", "stage13", "stage14", "stage15", "stage16", "stage17", "stage18", "stage19", "stage20", "stage21", "stage22", "stage23", "stage24" ], "use_mask_token": false }, "fusion_hidden_size": 256, "image_model_config": { "hidden_size": 1024, "image_size": 384, "model_type": "dinov2", "num_attention_heads": 16, "num_hidden_layers": 24, "out_features": [ "stage24" ], "out_indices": [ 24 ], "patch_size": 16, "stage_names": [ "stem", "stage1", "stage2", "stage3", "stage4", "stage5", "stage6", "stage7", "stage8", "stage9", "stage10", "stage11", "stage12", "stage13", "stage14", "stage15", "stage16", "stage17", "stage18", "stage19", "stage20", "stage21", "stage22", "stage23", "stage24" ], "use_mask_token": false }, "initializer_range": 0.02, "intermediate_feature_dims": [ 256, 256 ], "intermediate_hook_ids": [ 11, 5 ], "merge_padding_value": 3, "model_type": "depth_pro", "num_fov_head_layers": 2, "patch_model_config": { "hidden_size": 1024, "image_size": 384, "model_type": "dinov2", "num_attention_heads": 16, "num_hidden_layers": 24, "out_features": [ "stage24" ], "out_indices": [ 24 ], "patch_size": 16, "stage_names": [ "stem", "stage1", "stage2", "stage3", "stage4", "stage5", "stage6", "stage7", "stage8", "stage9", "stage10", "stage11", "stage12", "stage13", "stage14", "stage15", "stage16", "stage17", "stage18", "stage19", "stage20", "stage21", "stage22", "stage23", "stage24" ], "use_mask_token": false }, "patch_size": 384, "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": "4.49.0.dev0", "use_batch_norm_in_fusion_residual": false, "use_bias_in_fusion_residual": true, "use_fov_model": true }