{ "add_projection": false, "architectures": [ "ZoeDepthForDepthEstimation" ], "attractor_alpha": 1000, "attractor_gamma": 2, "attractor_kind": "mean", "backbone": null, "backbone_config": { "hidden_size": 1024, "image_size": 384, "intermediate_size": 4096, "model_type": "beit", "num_attention_heads": 16, "num_hidden_layers": 24, "out_features": [ "stage6", "stage12", "stage18", "stage24" ], "out_indices": [ 6, 12, 18, 24 ], "reshape_hidden_states": false, "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_relative_position_bias": true }, "backbone_hidden_size": 1024, "batch_norm_eps": 1e-05, "bin_centers_type": "softplus", "bin_configurations": [ { "max_depth": 10.0, "min_depth": 0.001, "n_bins": 64, "name": "nyu" } ], "bin_embedding_dim": 128, "bottleneck_features": 256, "fusion_hidden_size": 256, "head_in_index": -1, "hidden_act": "gelu", "initializer_range": 0.02, "max_temp": 50.0, "min_temp": 0.0212, "model_type": "zoedepth", "neck_hidden_sizes": [ 256, 512, 1024, 1024 ], "num_attractors": [ 16, 8, 4, 1 ], "num_patch_transformer_layers": null, "num_relative_features": 32, "patch_transformer_hidden_size": null, "patch_transformer_intermediate_size": null, "patch_transformer_num_attention_heads": null, "readout_type": "project", "reassemble_factors": [ 4, 2, 1, 0.5 ], "torch_dtype": "float32", "transformers_version": "4.42.0.dev0", "use_batch_norm_in_fusion_residual": false, "use_bias_in_fusion_residual": null, "use_pretrained_backbone": false }