search

IML-DKFZ / nnunet-workshop

42 stars 7 forks source link

About using Google Colab to run the nnUNet #1

Open Calvin-Pang opened 1 year ago

Calvin-Pang commented 1 year ago

Hi, I am a graduate student who is trying to use nnUNet in Colab.

I tried to run nnUNet preprocessing on the Liver dataset in Medical Decathlon Dataset with the Google Colab. But I found I will always met a freeze in the process even I have made the number of threads 1. So have you ever tried the liver tumor segmentation dataset in Colab? Or can you give me some advice about this issue?

Many thanks!

sten2lu commented 1 year ago

Hi @Calvin-Pang, I am sorry for the late reply.

I never ran preprocessing on the Liver Dataset of the Medical Decathlon Dataset on google colab. As freezing processes are often attributed to exceeding the RAM and some workers crashing, your idea to reduce the number of worker was correct. Perhaps you could try to run the same Code with nnU-Net Version 2, this Google Colab has now also been updated to nnU-Net version 2. So you might want to check it out again. The advantage of nnU-Net V2 is that it now throws errors when it gets stuck due to multiprocessing issues.

If you know exactly which type of configuration you want to prepare, you could also try the following command.

nnUNetv2_plan_and_preprocess -d <Liver Dataset> -c 3d_fullres -np 1 --verbose

Best regards, Carsten