disable optimizers setup during testing

[rohitgr7]

What does this PR do?

Disable configuring optimizers and schedulers during testing.

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 your PR does only one thing, instead of bundling different changes together? Otherwise, we ask you to create a separate PR for every change.
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?
• Did you verify new and existing tests pass locally with your changes?
• If you made a notable change (that affects users), did you update the CHANGELOG?

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 🙃

[pep8speaks]

Hello @rohitgr7! Thanks for updating this PR.

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

Comment last updated at 2020-09-28 21:20:45 UTC

[codecov]

Codecov Report

Merging #3059 into master will increase coverage by 5%.
The diff coverage is 93%.

@@           Coverage Diff           @@
##           master   #3059    +/-   ##
=======================================
+ Coverage      86%     91%    +5%     
=======================================
  Files         110     110            
  Lines        8220    8206    -14     
=======================================
+ Hits         7069    7463   +394     
+ Misses       1151     743   -408     

[rohitgr7]

amp tests are just checking that training is a success or not. There is no test to check whether the model and optimizers are actually configured with apex/native_amp or not.

[SkafteNicki]

@rohitgr7 PR is looking good, just wanted to ask why it is necessary to disable configuration during testing?

[rohitgr7]

@SkafteNicki if someone wants to use lightning module for the sole purpose of testing or monitoring the performance of their model, then only test part of lightning module is required so I don't think optimizers or schedulers are required here. If we keep it as it is, I might throw an error since .test actually calls .fit

[mergify]

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

[SkafteNicki]

@rohitgr7 good point. If there is a use case right now on master that produce an error (related to this), we should probably write an test that quantifies that this PR solves that :]

[rohitgr7]

@SkafteNicki a simple one I can think of is just implement test_step and test_epoch_end and call trainer.test(). Ideally it should work but it throws an error. We definitely need some tests for this. Marked as WIP.

[rohitgr7]

lots of refactoring going on with accelerators. I guess I'll wait for some time to work on this PR.

[rohitgr7]

should I write supporting tests for each backend?

[ananthsub]

I only have a few style comments, but the core looks good to me! Thank you for adding this @rohitgr7 !

[Borda]

seems some more GPU tests are failing, @rohitgr7 mind check it...

[rohitgr7]

@Borda fixed the bug.