Matthew Leigh
Matthew Leigh

Hi @awaelchli, Thanks for the response. You are right that the code I sent didn't quite show the bug. That was my mistake and was due to a typo. Instead...
I believe the issue is that Lightning is over relying on the length parameter to determine the end of the training epoch. If I were to write my own training...
Yeah I could give it a go. The only think that I think I need to account for is preventing the double execution of the valid epoch if the batch...