[bugfix] Reduce memory leaks

[tchaton]

What does this PR do?

This PR moves the optimizer states back to cpu on teardown and move ResultCollection Extra on cpu too.

Fixes #8463
Fixes #8430

Investigation for memory left:

from copy import _deepcopy_dispatch, deepcopy

def _deepcopy_dispatch_tensor(x, memo, deepcopy=deepcopy):
      print(x) 
      return x

_deepcopy_dispatch[torch.Tensor] = _deepcopy_dispatch_tensor

and use the deepcopy on Trainer, model

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

No.

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)
• Did you list all the breaking changes introduced by this pull request?

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 @tchaton! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

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

Comment last updated at 2021-07-21 08:49:52 UTC

[codecov]

Codecov Report

Merging #8490 (2719d03) into master (ea13f60) will decrease coverage by 4%.
The diff coverage is 84%.

@@           Coverage Diff           @@
##           master   #8490    +/-   ##
=======================================
- Coverage      92%     88%    -4%     
=======================================
  Files         217     217            
  Lines       14367   14390    +23     
=======================================
- Hits        13260   12669   -591     
- Misses       1107    1721   +614     

[kaushikb11]

PR breaking TPUs training. Looking into it.

[awaelchli]

@kaushikb11 here you are calling self.root_device anyway, despite the comment above?

[tchaton]

Yes, not sure to understand the reasoning too.