Skip to content
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

Jump to bottom

Enforce an epoch scheduler interval when using SWA #6588

Merged
merged 9 commits into from
Apr 6, 2021
Merged

Enforce an epoch scheduler interval when using SWA #6588

Changes from 1 commit
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
ee8b1a2
Update stochastic_weight_avg.py
mibaumgartner Mar 18, 2021
73ee646
add test
mibaumgartner Mar 23, 2021
2194981
remove blank lines
mibaumgartner Mar 23, 2021
6ddf3aa
remove class and add interval parameter
mibaumgartner Mar 23, 2021
fc257b2
Merge branch 'master' into bugfix/SWA_scheduler_step
carmocca Mar 23, 2021
51f2aba
CHANGELOG
carmocca Mar 23, 2021
424a09e
Merge branch 'master' into bugfix/SWA_scheduler_step
carmocca Mar 27, 2021
1581527
Merge branch 'master' into bugfix/SWA_scheduler_step
akihironitta Apr 2, 2021
e7f05a6
Merge branch 'master' into bugfix/SWA_scheduler_step
s-rog Apr 6, 2021
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
7 changes: 4 additions & 3 deletions pytorch_lightning/callbacks/stochastic_weight_avg.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,14 +187,15 @@ def on_train_epoch_start(self, trainer: 'pl.Trainer', pl_module: 'pl.LightningMo
anneal_strategy=self._annealing_strategy,
last_epoch=trainer.max_epochs if self._annealing_strategy == "cos" else -1
)
_scheduler_config = _get_default_scheduler_config()
assert _scheduler_config["interval"] == "epoch" and _scheduler_config["frequency"] == 1
_scheduler_config["scheduler"] = self._swa_scheduler

if trainer.lr_schedulers:
lr_scheduler = trainer.lr_schedulers[0]["scheduler"]
rank_zero_warn(f"Swapping lr_scheduler {lr_scheduler} for {self._swa_scheduler}")
trainer.lr_schedulers[0]["scheduler"] = self._swa_scheduler
trainer.lr_schedulers[0] = _scheduler_config
else:
_scheduler_config = _get_default_scheduler_config()
_scheduler_config["scheduler"] = self._swa_scheduler
trainer.lr_schedulers.append(_scheduler_config)

self.n_averaged = torch.tensor(0, dtype=torch.long, device=pl_module.device)
Expand Down