Closed thahmadalm closed 2 months ago
go to nnUNet/nnunetv2/training/nnUNetTrainer/nnUNetTrainer.py find self.num_epochs = 1000 change to 2000 if you like or use this nnUNetv2_train dataset_id 3d_fullres num_fold -tr nnUNetTrainer_2000epochs
I never understood why number of epochs had to be defined as seperate classes. It seems like it could easily be an cli argument.
yeah @osbm the second option is a cli command sorry if that wasn't clear
Use that command:
nnUNetv2_train 667 2d all -p nnUNetPlansSpineMini -device mps -tr nnUNetTrainer_10epochs --npz
(or 100, 50, ..., please look at the file that this parameter defined)
Tricky point !!! For only one epoch:
nnUNetv2_train 667 2d all -p nnUNetPlansSpineMini -device mps -tr nnUNetTrainer_1epoch --npz
hi fabian,
i am trying to run nnunet in google colab however i am confused how I can change the max number of epochs so the trainer doesn't run 1000 epochs. Any ideas?
I know you mentioned this "You need to create a new trainer class, derive from nnUNettrainerV2" but how can I do this?