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

overfit_pct now uses train loaders for val and test and does not shuffle #2214

Closed
wants to merge 1 commit into from
Closed

overfit_pct now uses train loaders for val and test and does not shuffle #2214

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
22 changes: 17 additions & 5 deletions pytorch_lightning/trainer/data_loading.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,17 +227,29 @@ def _reset_eval_dataloader(self, model: LightningModule, mode: str) -> Tuple[Uni
Returns:
Tuple (num_batches, dataloaders)
"""
dataloaders = self.request_dataloader(getattr(model, f'{mode}_dataloader'))
# use the training loader as val and test when overfitting
if self.overfit_pct > 0:
dataloaders = self.request_dataloader(getattr(model, f'train_dataloader'))
else:
dataloaders = self.request_dataloader(getattr(model, f'{mode}_dataloader'))

if not isinstance(dataloaders, list):
dataloaders = [dataloaders]

# shuffling in val and test set is bad practice
for loader in dataloaders:
# shuffling in val and test set is bad practice
if mode in ('val', 'test') and hasattr(loader, 'sampler') and isinstance(loader.sampler, RandomSampler):
rank_zero_warn(
f'Your {mode}_dataloader has shuffle=True, it is best practice to turn'
' this off for validation and test dataloaders.')

# when overfitting, remove the randomsampler for the train loaders
if self.overfit_pct > 0:
rank_zero_warn(
f'You requested to overfit but enabled training Dataloader shuffling. Disabling it for you'
)
loader.sampler = SequentialSampler(loader.dataset)
else:
rank_zero_warn(
f'Your {mode}_dataloader has shuffle=True, it is best practice to turn'
' this off for validation and test dataloaders.')

if any([dl is None for dl in dataloaders]):
rank_zero_warn("One of given dataloaders is None and it will be skipped.")
Expand Down