Early stopping when validation is disabled

[awaelchli]

Before submitting

• Was this discussed/approved via a Github issue? (no need for typos and docs improvements)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure to update the docs?
• Did you write any new necessary tests?
• If you made a notable change (that affects users), did you update the CHANGELOG?

What does this PR do?

Fixes #1201 (right now it's just just a check that the proposed fix by @Dunrar passes the unit tests).

• Early stopping falls back to training metrics returned in e.g. training_step
• Documented this behaviour
• Added test to make sure this doesn't get broken in the future

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

[awaelchli]

@jeremyjordan @neggert I think you guys worked on callbacks and early stopping. Is it ok to do this change here? Basically what we want (according to @williamFalcon) is that early stopping will work on the training metrics if no validation loop is defined (or it is disabled).
If you agree with this change, then I will add a unit test for the case I described.

[jeremyjordan]

@awaelchli yeah i think that looks fine!

[Borda]

LGTM 🚀

[awaelchli]

Thanks for looking at it already. I added a test and properly documented the behaviour. Could you review it again?

[codecov]

Codecov Report

Merging #1235 into master will increase coverage by 0%.
The diff coverage is 100%.

@@          Coverage Diff           @@
##           master   #1235   +/-   ##
======================================
  Coverage      92%     92%           
======================================
  Files          62      62           
  Lines        3181    3181           
======================================
+ Hits         2923    2924    +1     
+ Misses        258     257    -1     

[mergify]

Great job! =)