tutorial_train.py does not produce checkpoints

[jamesWalker55]

When running the tutorial_train.py script, it creates logs at lightning_logs. However, the lightning_logs\version_X\checkpoints folders are all empty:

Is this intended behaviour, or is there a bug preventing the checkpoints from saving?

[SwayStar123]

There is no code to save checkpoints. You need to make a checkpoint callback

[feiyangsuo]

from pytorch_lightning.callbacks import ModelCheckpoint
checkpoint_callback = ModelCheckpoint(
                dirpath={your work dir},
                every_n_train_steps={after how many gradient descent do save one model weights file},
                save_weights_only=False
            )

and replace callbacks=[logger] with callbacks=[logger, checkpoint_callback] in your trainer = pl.Trainer(... code parameters.

cuz one epoch may be too long to wait for, you may want to save weights after some certain iterations. This code might help.