[fix] Attach train+val dataloaders to trainer in trainer loop

[ananthsub]

What does this PR do?

Fixes #7208 .

The issue occurs because the validation dataloader is never attached to the trainer by the time we reach this stage in the training loop: https://github.com/PyTorchLightning/pytorch-lightning/blob/44d775fccfb825561937f6fa03fe258af25c2b83/pytorch_lightning/trainer/training_loop.py#L558-L560

Currently this is conditional based on this logic, which is called once at the start of training: https://github.com/PyTorchLightning/pytorch-lightning/blob/44d775fccfb825561937f6fa03fe258af25c2b83/pytorch_lightning/trainer/trainer.py#L604

Removing the reload_dataloaders_every_epoch check here gives me consistent behavior again.

I don't think this is a fully proper fix - we should be attaching these dataloaders to the trainer more clearly, and operating them on them for reloads later on in the training loop.

This is a behavior change that resulted after #6075. Before, run_evaluation ran, which populated the dataloader settings on the trainer, and then the checkpoint callback ran. After, the checkpoint callback ran first, before run_evaluation. the check for whether we run evaluation depends on the val dataloader being present, which was being set inside of run_evaluation.

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[codecov]

Codecov Report

Merging #7207 (c97d172) into master (338f5a3) will decrease coverage by 4%.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master   #7207    +/-   ##
=======================================
- Coverage      91%     87%    -4%     
=======================================
  Files         199     199            
  Lines       12797   12797            
=======================================
- Hits        11678   11145   -533     
- Misses       1119    1652   +533     

[pep8speaks]

Hello @ananthsub! Thanks for updating this PR.

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-04-29 22:07:06 UTC

[edenlightning]

@justusschock @awaelchli

[ananthsub]

@awaelchli @justusschock @tchaton what do you think?

[justusschock]

Looks good to me.

[ananthsub]

@tchaton @awaelchli @Borda mind taking a look?

[ananthsub]

@carmocca @Borda @tchaton mind taking a look?