Added serializability tests in all optimizers and fixed serializability bug in yogi

[ashutosh1919]

Related to #1294 .

@gabrieldemarmiesse - Please review this one.

[bot-of-gabrieldemarmiesse]

@SSaishruthi @manzilz @shreyashpatodia @PhilJd @pkan2 @junjiek @CyberZHG @lokhande-vishnu

You are owners of some files modified in this pull request.
Would you kindly review the changes whenever you have the time to?
Thank you very much.

[seanpmorgan]

Fixes look good, tyvm for the catch and much needed tests! Would you mind using the serialization test util for the comparisons though:
https://github.com/tensorflow/addons/blob/master/tensorflow_addons/testing/serialization.py#L53

In the future we could add a specialized check_optimizer_serialization as well (not needed for this PR).

[ashutosh1919]

@seanpmorgan , I think the method you suggested is too constrained.
I am checking this serialization.check_config(new_optimizer.get_config(), optimizer, False) and it is giving error as AttributeError: 'ConditionalGradient' object has no attribute '__name__'. What do it mean by that?

[ashutosh1919]

@seanpmorgan ,Please take a look at latest commit. With the use of check_config() in serialization.py, I have modified only one file and its tests - conditional_gradient.py and conditional_graident_test.py files. I have also slightly modified the function check_config(). Please review these three files and guide me whether I am doing right or not?

[seanpmorgan]

Hi @ashutosh1919 so the bazel tests are failing because you have to explicitly link python imports that are not in the same directory. See #1740 (just the optimizer BUILD) as a fix or we can merge master into this branch after that PR merges.

[googlebot]

All (the pull request submitter and all commit authors) CLAs are signed, but one or more commits were authored or co-authored by someone other than the pull request submitter.

We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that by leaving a comment that contains only @googlebot I consent. in this pull request.

Note to project maintainer: There may be cases where the author cannot leave a comment, or the comment is not properly detected as consent. In those cases, you can manually confirm consent of the commit author(s), and set the cla label to yes (if enabled on your project).

ℹ️ Googlers: Go here for more info.

[seanpmorgan]

@googlebot I consent.

[googlebot]

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

[seanpmorgan]

@seanpmorgan , I think the method you suggested is too constrained.
I am checking this serialization.check_config(new_optimizer.get_config(), optimizer, False) and it is giving error as AttributeError: 'ConditionalGradient' object has no attribute '__name__'. What do it mean by that?

Ah, this is a bit too restrictive. Could you revert your commits using the shared serialization check please. I'll create an issue to standardize all optimizer serialization checks, but want to get the bugfix in, using your first serialization check.

[ashutosh1919]

@seanpmorgan - I have reverted the changes. And please create a separate issue to add serailizability check for optimizers in serialization.py. I definitely want to work on it.

[seanpmorgan]

LGTM thanks for the fix! #1747 created for the next steps!