I am making use of the U-Mamba framework which is built upon nnunetv2.
I have Ubuntu 20.04 and Cuda 12.6 and followed all the instructions https://github.com/bowang-lab/U-Mamba and also tried building it from scratch, using newer pytorch versions etc. However, I always run into a multiprocessing error as on the image. I also tried first building nnunetv2 and then U-Mamba on top of it, but when running '''nnUNetv2_plan_and_preprocess -d 434 --verify_dataset_integrity''' I always get this runtime error.
Considering the dataset: It is MRI data with .nii.gz extension. My venv worked using just one channel. However, when I use two channels as data input, i get the runtime error. The dataset.json is as requested.
I also tried changing -npfp number of processes - which kind of changes where the runtime error occurs, but it still occurs.
Has anyone of you encountered similar problems or has advice whether there could be a package mismatch somewhere?
Hello,
I am making use of the U-Mamba framework which is built upon nnunetv2.
I have Ubuntu 20.04 and Cuda 12.6 and followed all the instructions https://github.com/bowang-lab/U-Mamba and also tried building it from scratch, using newer pytorch versions etc. However, I always run into a multiprocessing error as on the image. I also tried first building nnunetv2 and then U-Mamba on top of it, but when running '''nnUNetv2_plan_and_preprocess -d 434 --verify_dataset_integrity''' I always get this runtime error.
Considering the dataset: It is MRI data with .nii.gz extension. My venv worked using just one channel. However, when I use two channels as data input, i get the runtime error. The dataset.json is as requested.
I also tried changing -npfp number of processes - which kind of changes where the runtime error occurs, but it still occurs.
Has anyone of you encountered similar problems or has advice whether there could be a package mismatch somewhere?
I appreciate your help, thank you in advance!