low training speed on S3DIS?

[tangze5258]

every epoch took about 6000 s , somgthing wrong?

[tangze5258]

how to predict single *.nii file if the model trianing ends?

[saikat-roy]

Hey @tangze5258. Sorry for the late reply. You can use the following mednextv1_predict instead of nnUNet_predict as in the example here for inference on single files.

I have never trained on S3DIS but the training speeds depend heavily on the CPU, type of storage the training data is on (HDD, SSD etc) and the GPU used. If you could tell me about your configuration, I could confirm whether this indeed is the issue.