FabricModule: wrap forward methods instead of monkeypatch-based redirect

[tonyf]

What does this PR do?

When running with torch.compile and ModelParallelStrategy the monkeypatch method of redirecting forward methods through the FabricModule's custom forward method raises an exception if the marked method does not have the same signature as the original forward method.

Instead of monkeypatching forward to the method being called, this PR wraps the function to ensure the right strategy contexts are provided.

This could be cleaned up further by just wrapping the functions in mark_forward_method but that would require touching _LIGHTNING_MODULE_STEP_METHODS = ("training_step", "validation_step", "test_step", "predict_step").

Fixes #20710

Before submitting

• Was this discussed/agreed via a GitHub issue? (not for typos and docs)
• [x ] 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:

Reviewer checklist

• 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

---

📚 Documentation preview 📚: https://pytorch-lightning--20711.org.readthedocs.build/en/20711/

[Borda]

seems the falling stand-alone test is a false alarm :)
but the others seems real...

[codecov]

Codecov Report

❌ Patch coverage is 8.33333% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 81%. Comparing base (2676332) to head (2f5cf7b).
⚠️ Report is 2 commits behind head on master.

❗ There is a different number of reports uploaded between BASE (2676332) and HEAD (2f5cf7b). Click for more details.

HEAD has 104 uploads less than BASE

Flag	BASE (2676332)	HEAD (2f5cf7b)
cpu	53	27
lightning	31	16
pytest	29	3
python3.11	12	6
python	5	3
lightning_fabric	13	2
python3.10	12	6
python3.12.7	18	9
python3.12	6	3

Additional details and impacted files

@@            Coverage Diff            @@
##           master   #20711     +/-   ##
=========================================
- Coverage      87%      81%     -6%     
=========================================
  Files         268      268             
  Lines       23354    23354             
=========================================
- Hits        20346    18913   -1433     
- Misses       3008     4441   +1433     

[tonyf]

Ah! looks like I need to update the redirection tests. Will try to get around to it soon. Currently using this fix in prod successfully

[stale]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. If you need further help see our docs: https://lightning.ai/docs/pytorch/latest/generated/CONTRIBUTING.html#pull-request or ask the assistance of a core contributor here or on Discord. Thank you for your contributions.

[deependujha]

Hi @tonyf ,
Could you please take a look and fix failing tests so we can proceed with the merge?