Removing unecessary early stopping calls

[elkotito]

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 #1751

PR review

The removed early stopping calls are conditioned in different cases.

• The first one is called right after the validation dataset evaluation.
• The second one is in case of no validation dataset i.e. no validation_step implemented.

It's enough to call it once after the training because self.early_stop_callback.on_epoch_end can read required metrics from trainer.callback_metrics.

[williamFalcon]

i think there was a reason for both of those. need to think about this more

[jeremyjordan]

@williamFalcon Directly calling the callback methods (eg. on_epoch_end) is an anti-pattern, we should delegate that responsibility to the callback handler. When a user is writing callback code for on_epoch_end they should be able to trust that it will only be called at the end of an epoch.

[awaelchli]

It looks like this also fixes an issue with "patience".
If so I would also list it in the changelog that patience is fixed.

[mergify]

This pull request is now in conflict... :(

[codecov]

Codecov Report

Merging #1863 into master will decrease coverage by 0%.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           master   #1863   +/-   ##
======================================
- Coverage      88%     88%   -0%     
======================================
  Files          74      74           
  Lines        4650    4645    -5     
======================================
- Hits         4070    4065    -5     
  Misses        580     580           

[mergify]

This pull request is now in conflict... :(

[mergify]

This pull request is now in conflict... :(