fix missing call to untoggle_optimizer when accumulating gradients

[awaelchli]

What does this PR do?

Fixes #8281

The training loop toggles the optimizers in case there are mutliple, and untoggles after the optimizer step is completed.
However, the untoggle is missing during the accumulation phase, and it is certainly necessary.

Every toggle_optimizer() call needs a matching untoggle_optimizer() call.

Test fails on master

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)

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?

I made sure I had fun coding 🙃

[codecov]

Codecov Report

Merging #8284 (43c7fb0) into master (ea5cfd2) will decrease coverage by 5%.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master   #8284    +/-   ##
=======================================
- Coverage      93%     88%    -5%     
=======================================
  Files         212     212            
  Lines       13716   13729    +13     
=======================================
- Hits        12747   12075   -672     
- Misses        969    1654   +685     

[carmocca]

Is the milestone correct? Is this also an issue in the bug-fix branch or just after the loop refactor?

edit: nvm saw the original reported issue is for 1.3.7

[bmahlbrand]

I know this is closed, but I just pulled master, and tried testing w/3 optimizers instead of 2 (this problem is fixed when I only have 2) and the issue persists.

Check this out: #8365

[bmahlbrand]

This assumes len() == 2, consider 3 optimizers - it should be a cyclic toggle instead of boolean