qwen gbs 2x

[malay-nagda]

What does this PR do ?

Updated GBS for Qwen3 30B_a3B H100

Changelog

global_batch_size=1024,

GitHub Actions CI

See the CI sectionin 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

• Related to # (issue)

Summary by CodeRabbit

• Chores
  • Updated batch size configuration for QWEN 30B model training on H100 GPUs to enhance computational efficiency across BF16 and FP8 precision modes.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[coderabbitai]

📝 Walkthrough

Walkthrough

This change increases the global_batch_size parameter to 1024 in two QWEN3 pretraining configurations for H100 GPUs (BF16 and FP8 variants) by adding the parameter to their respective configuration calls.

Changes

Cohort / File(s)	Summary
QWEN3 Configuration Updates scripts/performance/configs/qwen/qwen3_workload_base_configs.py	Increased global_batch_size to 1024 for QWEN3_30B_A3B_PRETRAIN_CONFIG_H100_BF16_V1 and QWEN3_30B_A3B_PRETRAIN_CONFIG_H100_FP8_CS_V1 configurations.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Suggested labels

performance, r0.3.0

Suggested reviewers

• ko3n1g
• erhoo82

🚥 Pre-merge checks | ✅ 2 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Test Results For Major Changes	⚠️ Warning	PR doubles global batch size for H100 Qwen3 configs but provides no performance numbers, convergence analysis, or regression testing in the description.	Add before-and-after performance numbers, convergence analysis, and validation results to the PR description to demonstrate the change is safe.
Title check	❓ Inconclusive	The title 'qwen gbs 2x' is vague and uses non-descriptive abbreviations that lack meaningful context about the changeset.	Use a clearer title such as 'Increase global batch size to 1024 for QWEN3 H100 configs' to better convey the specific change being made.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ 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 malay/qwen_gbs_2x

---

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.}

[ko3n1g]

Can you share the pipeline run of the test with RC4 of the affected configs?