Training Stopped

[ThisIsForReview]

Also the training stopped after 2 epoches with the error
FileNotFoundError: [Errno 2] No such file or directory: "lightning_logs\'mnist'\version_1\checkpoints\last.ckpt"

Not sure why last.ckpt was not saved?

Thanks

JB

[fdraxler]

Hi, thanks for your question. Can you provide more details for which command you executed so that we can reproduce the issue? Thanks!

[ThisIsForReview]

I found every_n_epochs: 1 and save_top_k: 2 can work, but every_n_epochs > 1 does not work. It seems when every_n_epochs > 1, no last.ckpt has been saved.

JB

[fdraxler]

So you are referring to the model checkpointing, right? What behavior are you trying to achieve?

[ThisIsForReview]

It seems this is an issue from pytorch lightning. When save_last = True, every_n_epochs cannot be larger than 1. Lightning-trainable sets save_last = True, so every_n_epochs = 5 does not work.

JB