Fix LR scheduler behaviour with AMP

[milesial]

What does this PR do?

When training when native AMP and a LR scheduler, we get this warning that indicates that a LR step has been taken when an optimizer step was skipped (expected at the beginning of the training with native AMP):

/usr/local/lib/python3.8/dist-packages/torch/optim/lr_scheduler.py:138: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`.  Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate

Fixes #16228 #5558

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 list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the 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

[milesial]

In the process of fixing tests I discovered and fixed a bug where the scheduler wouldn't match its optimizer when multiple optimizers are instantiated with frequencies. Now the optimizers and schedulers match and alternate as they should, resetting the cycle every epoch.

[milesial]

@carmocca Ready for final review

[tchaton]

Could you add a test to verify this works properly ?

[milesial]

I modified the third case of test_step_scheduling_for_multiple_optimizers_with_frequency so that it tests that

[carmocca]

Can you check the failing tests?

[carmocca]

I added this line so that our CI fails if this warning appears. This way it tests that your patch works as expected.

[milesial]

Thanks, but this also makes IPU tests fail, this PR is focused on GPU. Not sure where to fix the issue on IPUs

[akihironitta]

@milesial IMHO, we could just skip the IPU test failures as done in my old PR #11755 for now.

[milesial]

The way I fixed the tests/tests_pytorch/models/test_hooks.py::test_trainer_model_hook_system_fit[True-kwargs1] tests is very flaky, so I'd appreciate if someone more familiar with these tests comes up with a better fix.
edit: seems like it didn't even fix it...

[milesial]

I also modified native_amp.py in both lightning_fabric and pytorch_lightning. It doesn't seem like the lightning_fabric one is called for a typical workflow, so I don't know if it's the right way

[mrembalski]

Hi @Borda, I also encounter the same issue. Will this be merged?

[Borda]

Hi @Borda, I also encounter the same issue. Will this be merged?

Let me check what is missing here...

[alimoezzi]

Is this PR merged already? I'm still having this issue.

[Borda]

there were some failing tests, @milesial mind have a look?

[gitguardian]

⚠️ GitGuardian has uncovered 2 secrets following the scan of your pull request.

Please consider investigating the findings and remediating the incidents. Failure to do so may lead to compromising the associated services or software components.

🔎 Detected hardcoded secrets in your pull request

GitGuardian id	GitGuardian status	Secret	Commit	Filename
-		Generic High Entropy Secret	78fa3af	tests/tests_app/utilities/test_login.py	View secret
-		Base64 Basic Authentication	78fa3af	tests/tests_app/utilities/test_login.py	View secret

🛠 Guidelines to remediate hardcoded secrets

1. Understand the implications of revoking this secret by investigating where it is used in your code.
2. Replace and store your secret safely. Learn here the best practices.
3. Revoke and rotate this secret.
4. If possible, rewrite git history. Rewriting git history is not a trivial act. You might completely break other contributing developers' workflow and you risk accidentally deleting legitimate data.

To avoid such incidents in the future consider

• following these best practices for managing and storing secrets including API keys and other credentials
• install secret detection on pre-commit to catch secret before it leaves your machine and ease remediation.

---

<sup>🦉 GitGuardian detects secrets in your source code to help developers and security teams secure the modern development process. You are seeing this because you or someone else with access to this repository has authorized GitGuardian to scan your pull request.

Our GitHub checks need improvements? Share your feedbacks!</sup>

[codecov]

Codecov Report

Merging #16229 (965fc03) into master (6497e36) will decrease coverage by 54%.
Report is 1 commits behind head on master.
The diff coverage is 29%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #16229      +/-   ##
==========================================
- Coverage      83%      29%     -54%     
==========================================
  Files         450      442       -8     
  Lines       38089    37941     -148     
==========================================
- Hits        31803    11015   -20788     
- Misses       6286    26926   +20640