[ci] Refactor Gemma3-VL launch script to run finetune and packed tests separately

[cuichenx]

What does this PR do?

Refactor the Gemma3-VL recipe functional test launch script to run test_gemma3_vl_finetune_recipes and test_gemma3_vl_finetune_packed_recipes as separate torch.distributed.run invocations for better test isolation.

Changelog

• Split single torch.distributed.run invocation in L2_Launch_recipes_gemma_vl.sh into two separate launches using pytest -k to select each test individually

GitHub Actions CI

See the CI section in the Contributing doc for how to trigger the CI.
A Nvidia developer will need to approve and trigger the CI for external contributors.

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
  • Reviewer: Does the PR have correct import guards for all optional libraries?

If you haven't finished some of the above items you can still open "Draft" PR.

Additional Information

N/A

Made with Cursor

Summary by CodeRabbit

• Tests
  • Added test coverage for Gemma3-VL finetune recipes
  • Improved test filtering for finetune packed recipes validation

[coderabbitai]

📝 Walkthrough

Walkthrough

A test script is updated to add a new pytest execution block for Gemma3-VL finetune recipes and refine the existing finetune packed recipes test to use a specific test filter instead of running all tests in the file.

Changes

Cohort / File(s)	Summary
Gemma VL Test Script tests/functional_tests/L2_Launch_recipes_gemma_vl.sh	Adds new test execution block for Gemma3-VL finetune recipes with pytest and coverage reporting; modifies existing finetune packed recipes test invocation to target a specific test selection rather than all tests in the file.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~5 minutes

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Test Results For Major Changes	✅ Passed	PR refactors test launch infrastructure by splitting a single torch.distributed.run invocation into two separate ones for improved test isolation, without affecting model code, numerics, or convergence—qualifying as a minor change.
Title check	✅ Passed	The title accurately describes the main change: refactoring the Gemma3-VL launch script to separate finetune and packed recipe tests into independent test executions.

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

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch chcui/refactor-gemma-vl-launch-script

---

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.}

[theothermike]

tested and confirmed it resolves the segfault in the test