fix(ci): recover from corrupted MMMU parquet cache

[harvenstar]

When MMMU dataset parquet files are corrupted in HuggingFace cache (ArrowInvalid: Parquet magic bytes not found), the lmms_eval process fails without producing JSON results, causing CI test failures.
https://github.com/sgl-project/sglang/actions/runs/21076101101/job/60658633322?pr=16388
https://github.com/sgl-project/sglang/actions/runs/21075841576/job/60653165682?pr=16561

This fix adds automatic recovery in mmmu_vlm_kit.py:

• Detects parquet corruption errors in lmms_eval subprocess output
• Cleans up the corrupted MMMU dataset cache directory (prioritizes CI path /hf_home, fallback to HF_HOME)
• Retries once with HF_HUB_OFFLINE=0 and force_redownload mode
• Only affects MMMU parquet corruption cases (minimal impact)

Fixes the intermittent CI failures on MMMU evaluation tests.

Motivation

Modifications

Accuracy Tests

Benchmarking and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.

Review Process

1. Ping Merge Oncalls to start the PR flow. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • /tag-run-ci-label, /rerun-failed-ci, /tag-and-rerun-ci
4. After green CI and required approvals, ask Merge Oncalls to merge.

[gemini-code-assist]

Summary of Changes

Hello @harvenstar, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly improves the robustness of the continuous integration pipeline by introducing an automatic recovery mechanism for corrupted HuggingFace dataset caches, specifically for MMMU evaluations. By detecting data corruption, clearing the problematic cache, and retrying the evaluation with a forced redownload, the system can now self-heal from transient data issues, leading to more stable and reliable CI test results.

Highlights

• Automatic Error Detection: Implemented a mechanism to automatically detect ArrowInvalid: Parquet magic bytes not found errors, specifically targeting corrupted MMMU dataset parquet files during lmms_eval subprocess execution.
• Corrupted Cache Cleanup: Added functionality to clean up the corrupted MMMU dataset cache directory. This prioritizes the CI path (/hf_home) and falls back to the standard HuggingFace cache location (HF_HOME or ~/.cache/huggingface).
• Retry Mechanism: Introduced a retry logic for lmms_eval commands. Upon detecting and cleaning a corrupted cache, the evaluation is retried once with HF_HUB_OFFLINE=0 and HF_DATASETS_DOWNLOAD_MODE=force_redownload to ensure a fresh download of the dataset.
• CI Stability Improvement: This fix directly addresses intermittent CI failures caused by corrupted MMMU evaluation datasets, enhancing the reliability of continuous integration tests.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces a robust retry mechanism to handle intermittent CI failures caused by corrupted MMMU dataset parquet files. The changes are well-targeted and introduce helper functions to detect the specific error, clean up the cache, and retry the evaluation. My review focuses on improving the logging and error reporting within the new retry logic to ensure that subprocess outputs are not lost, which is crucial for debugging.

[b8zhong]

/tag-and-rerun-ci