fix _no_split_modules subscript error for transformers v5

[Maxusmusti]

Summary

Fixes TypeError: 'set' object is not subscriptable when running FSDP training with transformers v5.x.

In transformers v5, _no_split_modules is now converted to a set during model initialization (post_init()), whereas it was previously a list. The code in get_fsdp_config() was using [0] to access the first element, which fails on sets.

Fix: Use next(iter(...)) to retrieve an element, which works with both sets and lists.

Test plan

• Verified fix resolves the error in SFT validation script
• Updated unit test mock to use a set to match transformers v5 behavior

Summary by CodeRabbit

• Bug Fixes
  • Improved robustness of the FSDP wrap policy configuration during model acceleration.

[coderabbitai]

📝 Walkthrough

Walkthrough

The pull request refactors container access patterns by replacing direct index-based retrieval with iterator-based retrieval for _no_split_modules, while updating the corresponding test fixture to use a set instead of a list, ensuring consistency between implementation and test.

Changes

Cohort / File(s)	Summary
FSDP wrap policy accessor refactor src/instructlab/training/accelerator.py	Replaced direct index access [0] with safe iterator-based retrieval next(iter(...)) for _no_split_modules to support non-list iterables.
Mock fixture update tests/unit/test_accelerator.py	Changed _no_split_modules from list to set in mock_model fixture to align with refactored iterator-based accessor logic.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Suggested labels

one-approval

Poem

🐰 From indexed lists we hop away,
To iterators light and fleet!
Sets replace the ordered array,
A refactoring so neat!
Safer access, cleaner flow,
Now the code can better grow! ✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: fixing a subscript error with _no_split_modules that occurs in transformers v5 due to the attribute being a set rather than a list.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/transformers-v5-no-split-modules-set

---

No actionable comments were generated in the recent review. 🎉

Tip

Issue Planner is now in beta. Read the docs and try it out! Share your feedback on Discord.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[RobotSail]

LGTM