Cannot Reproduce BraTS Challenge 2020 Results

[nathanjbeaumont]

Hello,
I am trying to run the pretrained nnUNet model for glioma segmentation which won BraTS 2020. I installed Docker Desktop, then downloaded the docker image at fabianisensee/isen2020, then downloaded wsl 2 for Windows 11 and checked "Use the WSL 2 based engine" in Docker Desktop. Based on the instructions here I ran the container with the command:
docker run --rm --gpus=all -v="C:\Grad School\Research\BraTS\data\MICCAI_BraTS2020_ValidationData\BraTS20_Validation_001":/data -it fabianisensee/isen2020

And nothing was output to the results folder.

In the terminal the only output was:

WARNING: The NVIDIA Driver was not detected.  GPU functionality will not be available.
   Use 'nvidia-docker run' to start this container; see
   https://github.com/NVIDIA/nvidia-docker/wiki/nvidia-docker .

NOTE: MOFED driver for multi-node communication was not detected.
      Multi-node communication performance may be reduced.

NOTE: The SHMEM allocation limit is set to the default of 64MB.  This may be
   insufficient for PyTorch.  NVIDIA recommends the use of the following flags:
   nvidia-docker run --ipc=host ...

Even though my gpu is visible when I run nvidia-smi in the image, but inference should work with only the CPU anyway right?

This is what my data folder looks like, it is a single patient in the BraTS 2020 validation dataset:

Any advice on how to get segmentations written into the results folder?

[nathanjbeaumont]

After looking through the dockerfile history I found the model and its parameters are saved in the folder /params.

I'm going to see if I can get the model running with the files in /params

[nathanjbeaumont]

After more digging I've found the model classes in /opt/conda/lib/python3.6/site-packages/nnunet/training/network_training/BraTS_trainer/nnUNetTrainerV2BraTSRegions.py, and it seems /opt/conda/lib/python3.6/site-packages/nnunet/dataset_conversion/Task082_BraTS_2020.py is where dataset conversion is done, however this has hardcoded data directory strings.

If someone could tell me what commands need to be run in bash to get inference working on the input folder of .nii.gz files it would be greatly appreciated.