Make unimplemented dataloader hooks raise NotImplementedError

[Tshimanga]

What does this PR do?

Fixes #8752

Does your PR introduce any breaking changes? If yes, please list them.

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 list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is welcome to review the PR.
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? yes

Make sure you had fun coding 🙃

@ananthsub I added the raise NotImplementedError on the other dataloader hooks, but I'm still curious about that choice if you have a moment

@ananthsub so the ticket states This applies for all the dataloader hooks in DataHooks, not just train_dataloader, but based on the docstrings on all the other dataloader hooks {val, test, predict}_dataloader are intentionally optional hooks whose presence is not necessary to run the Trainer. Is the assertion here that users will be required to override all of these dataloader hooks?

[ananthsub]

thanks for working on this @Tshimanga !
Please update the PR title to be more descriptive of the actual change

[ananthsub]

This applies for all the dataloader hooks in DataHooks, not just train_dataloader, but based on the docstrings on all the other dataloader hooks {val, test, predict}_dataloader are intentionally optional hooks whose presence is not necessary to run the Trainer. Is the assertion here that users will be required to override all of these dataloader hooks?

Technically, all of the hooks are optional to implement, depending on which Trainer function is called. For instance, if you call Trainer.validate the train dataloader is never checked or invoked. Additionally, the Trainer checks whether the corresponding dataloader hook is overridden before that loop's execution. However, if that override check fails for any reason, this will let us know sooner.

[codecov]

Codecov Report

Merging #9161 (95b8a9c) into master (0dfc6a1) will decrease coverage by 4%.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master   #9161    +/-   ##
=======================================
- Coverage      92%     88%    -4%     
=======================================
  Files         176     176            
  Lines       14857   14858     +1     
=======================================
- Hits        13703   13085   -618     
- Misses       1154    1773   +619     

[tchaton]

LGTM !

[Tshimanga]

@ananthsub @tchaton @awaelchli @carmocca
It looks like the required test PL.pytorch-lightning (GPUs) is hanging on pl_examples/test_examples.py::test_examples_cpu, but there isn't an error indicated. It just fails to progress and gets cancelled. I've looked at the test and I think I lack familiarity to pick out how/if my changes might have caused this. Some assistance would be appreciated 🙏🏽

https://dev.azure.com/PytorchLightning/pytorch-lightning/_build/results?buildId=27719&view=logs&j=48e2c176-e5bf-5c04-26fb-02608a8a8927&t=fba91523-81e0-5703-785d-4a83eb9c4946&l=26

[awaelchli]

it's not from your PR, pretty sure
related #9021

[Tshimanga]

@awaelchli hmm, looks like another failure. maybe #9021 just needs to be merged first?

[Tshimanga]

@awaelchli hmm it looks like generally those tests do creep pretty close to the 45min upper limit though