Skip to content

Fix EMA callback: resume training from the raw weights, not the EMA a… #785

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
sundusaijaz merged 1 commit into from
Jul 20, 2026
Merged

Fix EMA callback: resume training from the raw weights, not the EMA a… #785

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions src/schnetpack/train/callbacks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,6 +125,10 @@ def on_fit_start(self, trainer, pl_module: AtomisticTask):
if self._to_load is not None:
self.ema.load_state_dict(self._to_load)
self._to_load = None
if self.ema.collected_params is not None:
# checkpoints hold the EMA average (saved during validation);
# restore the raw training weights before store()/copy_to()
self.ema.restore()

# load average parameters, to have same starting point as after validation
self.ema.store()
Expand All @@ -141,6 +145,10 @@ def on_train_batch_end(self, trainer, pl_module: AtomisticTask, *args, **kwargs)
def on_validation_epoch_start(
self, trainer: "pl.Trainer", pl_module: AtomisticTask, *args, **kwargs
):
if trainer.sanity_checking:
# on_fit_start already stored the raw weights and copied the EMA
# average in; storing again would overwrite them with the average
return
self.ema.store()
self.ema.copy_to()

Expand Down
Loading