[WIP] Stricter argument hygiene for passing datamodules to trainer.fit or t…

[ananthsub]

What does this PR do?

• Deprecates passing datamodules as an unnamed argument to trainer.fit and trainer.tune
• Provides consistency with how datamodules are passed to validate and test
• Allows stronger typing for train_dataloader in these APIs

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 🙃

[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 2021-05-04 06:44:45 UTC

[codecov]

Codecov Report

Merging #7329 (4ed2254) into master (597b309) will increase coverage by 0%.
The diff coverage is 100%.

@@          Coverage Diff           @@
##           master   #7329   +/-   ##
======================================
  Coverage      87%     87%           
======================================
  Files         200     200           
  Lines       12896   12899    +3     
======================================
+ Hits        11204   11231   +27     
+ Misses       1692    1668   -24     

[carmocca]

I actually have a WIP (waiting for 1.4) branch to do exactly the opposite - adding the positional datamodule for the other functions.

This is a very widely used pattern:

trainer.fit(model, dm)

I'm not sure stronger typing is enough reason to remove it.

This was also asked by @awaelchli here: #7258 (comment)

[Borda]

nice :]

[Borda]

@ananthsub @carmocca just thinking if we shall define this kind of collection of messages which are repetitive in multiple places on the codebase and use them as constant?

[tchaton]

LGTM ! Thanks for this.

[s-rog]

WIP or ready to go? 😅
Either way, LGTM!

[tchaton]

Waiting on Carlos resolution. I agree it is a common use-case and will frustrate users.

[awaelchli]

Not sure about this change. I understand the motivation but I don't think it is necessary. When the user will upgrade from 1.2 to 1.3, they will already see plenty of deprecation messages and this one here seems unnecessary. I vote for holding back on this.

[awaelchli]

Do we go with #7431 or vote for one of them?

[carmocca]

IMO this is a no-go considering it's too ingrained into user's brains

[awaelchli]

What's the verdict on this one? 😅 we have accepts and we have some arguments against. A vote with core?

[tchaton]

Yes, validate Carlos call there. This is too much of a common usage to be dropped.