[pyTorch/nnUnet] TypeError: Illegal pipeline output type. The output 0 contains a nested `DataNode`. Missing list/tuple expansion (*) is the likely cause.

[chuxiang93]

Related to [pyTorch/nnUnet]

Describe the bug When I run "python scripts/inference.py --data /data/01_3d/val --dim 3 --fold 0 --ckpt_path <path/to/checkpoint> --amp --tta --save_preds", the program reported a bug "TypeError: Illegal pipeline output type. The output 0 contains a nested DataNode. Missing list/tuple expansion (*) is the likely cause."

Data folder:

To Reproduce Steps to reproduce the behavior: 1, python preprocess.py --task 01 --dim 3 --exec_mode val 2, python scripts/inference.py --data /data/01_3d/val --dim 3 --fold 0 --ckpt_path <path/to/checkpoint> --amp --tta --save_preds

Environment Please provide at least:

• Container version: pytorch 1.7
• GPUs in the system: V100
• CUDA driver version: 460.91

[michal2409]

Thanks for spotting that! I will push code with fix soon.

In the meantime you can fix it locally by replacing this line to the self.input_y = self.get_reader(kwargs["lbls"]) if kwargs["lbls"] is not None else None

[michal2409]

@nv-kkudrynski this is already fixed.