UnboundLocalError: local variable 'any_on_epoch' referenced before assignment in prediction loop

[farhadrgh]

Bug description

UnboundLocalError raises when using the predict method with return_predictions=False.

This is due to any_on_epoch not being defined if data_fetcher is an instance of _DataLoaderIterDataFetcher.

What version are you seeing the problem on?

v2.4

How to reproduce the bug

No response

Error messages and logs

UnboundLocalError: local variable 'any_on_epoch' referenced before assignment

Environment

Current environment

#- PyTorch Lightning Version (e.g., 2.4.0):
#- PyTorch Version (e.g., 2.4):
#- Python version (e.g., 3.12):
#- OS (e.g., Linux):
#- CUDA/cuDNN version:
#- GPU models and configuration:
#- How you installed Lightning(`conda`, `pip`, source):

More info

No response

[lantiga]

nice catch

This is due to any_on_epoch not being defined if data_fetcher is not an instance of _DataLoaderIterDataFetcher.

it's actually the other way around, it errors out when it is an instance of _DataLoaderIterDataFetcher

the solution is to replace lines 236 and 237:

pytorch-lightning/src/lightning/pytorch/loops/prediction_loop.py

Lines 236 to 237 in be608fa

	if not using_dataloader_iter:
	any_on_epoch = self._store_data_for_prediction_writer(batch_idx, dataloader_idx)

with

any_on_epoch = self._store_data_for_prediction_writer(batch_idx, dataloader_idx) if not using_dataloader_iter else False

Would you like to volunteer for a quick PR?

[farhadrgh]

Thanks, Yes, I just fixed the issue description as well.