{ "env": { "seed": 42, "verbose": true }, "accelerator": { "report_to": ["tensorboard"], "mixed_precision": "fp16", "accumulation_nsteps": 4, "timeout": 36000 }, "inference": { "images": "assets/demos", "masks": "assets/masks", "min_pixels": 580000, "max_pixels": 620000 }, "evaluation": { "alignment": "median", "datasets_dir": "/home/haodong/data/DA-2/Evaluation", "metric_names": [ "abs_relative_difference", "squared_relative_difference", "rmse_linear", "rmse_log", "log10", "delta1_acc", "delta2_acc", "delta3_acc", "i_rmse", "silog_rmse" ], "metric_show": { "abs_relative_difference": "AbsRel", "delta1_acc": "δ_1" }, "datasets": { "2d3ds": { "dir": "2D3DS/wo_xyz", "filenames": "eval/datasets/splits/2d3ds.txt", "alignment_max_res": 2048 }, "matterport3d": { "dir": "Matterport3D", "filenames": "eval/datasets/splits/matterport3d.txt", "alignment_max_res": 2048 }, "panosuncg": { "dir": "PanoSUNCG/rotated", "filenames": "eval/datasets/splits/panosuncg.txt", "alignment_max_res": 1024 } } }, "spherevit": { "vit_w_esphere": { "input_dims": [1024, 1024, 1024, 1024], "hidden_dim": 512, "num_heads": 8, "expansion": 4, "num_layers_head": [2, 2, 2], "dropout": 0.0, "layer_scale": 0.0001, "out_dim": 64, "kernel_size": 3, "num_prompt_blocks": 1, "use_norm": false }, "sphere": { "width": 1092, "height": 546, "hfov": 6.2832, "vfov": 3.1416 } } }