Specify version of data-multi-subject to use

[melisende-st-amour]

I tried running the training script for contrast-agnostic using this command: bash contrast-agnostic-softseg-spinalcord/scripts/train_contrast_agnostic.sh

I ran into this error: File "/home/ge.polymtl.ca/mestaa/contrast-agnostic-softseg-spinalcord/nnUnet/02_create_msd_data.py", line 504, in

I checked the contrast-agnostic-softseg-spinalcord/nnUnet/02_create_msd_data.py script and found that these lines are specific to an older version of data-multi-subject where the preprocessed data was in data-multi-subject/derivatives/data_preprocessed

contrast-agnostic-softseg-spinalcord/nnUnet/02_create_msd_data.py

Lines 142 to 144 in bfcb835

	if dataset_name == 'data-multi-subject':
	# get only the (preprocessed) subject files, which are in the `derivatives` folder
	path_files = os.path.join(dataset_path, 'derivatives', 'data_preprocessed', 'sub-*', '**', f'*.nii.gz')

The most recent version of data-multi-subject doesn't have data_preprocessed.

[naga-karthik]

hey @melisende-st-amour, thanks for opening the issue! there was a bigger problem about dataset versioning which I had to fix. Can you please checkout to my branch (i.e. do git checkout nk/clean-repo and git pull) and try running the script again?

[naga-karthik]

subjects_to_include is not clear