How to get nnUNetv2_find_best_configuration to work? (problems with fold) #1677
-
|
Hello everyone, thanks in advance for the help. I trained my dataset like this
Then I wanted to use the nnUNetv2_find_best_configuration function to get inference commands. However, I get an error declaring that I have not trained fold 0. It does not recognize the "all" type of training. Do I have to repeat 5 different trainings, one for each fold? Or at least train fold 0 to use the nnUNetv2_find_best_configuration ? Thanks for your help. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Hi eziolanza. I think nnUNetv2_find_best_configuration requires you to have trained more than one fold and/or more than one configuration. You trained one configuration and one fold -- 'all' doesn't mean nnU-Net will train all folds, rather than nnU-Net will use all training images for training (effectively eliminating any possible real folds) -- so there are is no best configuration given that there is only one configuration. I hope I have made sense. Vicent |
Beta Was this translation helpful? Give feedback.
Hi eziolanza.
I think nnUNetv2_find_best_configuration requires you to have trained more than one fold and/or more than one configuration. You trained one configuration and one fold -- 'all' doesn't mean nnU-Net will train all folds, rather than nnU-Net will use all training images for training (effectively eliminating any possible real folds) -- so there are is no best configuration given that there is only one configuration. I hope I have made sense.
Vicent