Open Pascal-bucketbreaker opened 3 days ago
A simple solution for the checkpointing is to change it in nnUNetTrainer.py
### checkpoint saving stuff
self.save_every = 50
self.disable_checkpointing = False
You could also potentially create a custom trainer for this as well as they do with epochs.
For validation, the validation cases files can be found in the result. Even when you complete an entire training cycle, it shows which cases are in validation.
And also, the programme will pause running for unknown reason until you press the "Enter" key. Besides, why I added the system variable in the windows control panel, but I can't find them after running the 'set' command, the programme also can't find it and get the error.
I have seen your replies. Thanks for your help!
Save a checkpoint after 50 epochs is too long. How can I change it? And, how can I get each fold of validation data, including its original image, prediction, and mask? I want them for further analysis.