[Bugfix] Fix OOM tests in initialization test

[Isotr0py]

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

Purpose

• Fix skipped OOM tests in tests/models/test_initialization.py

Test Plan

pytest -s -v tests/models/test_initialization.py

Test Result

• Previous three OOM tests should pass now

(Optional) Documentation Update

[gemini-code-assist]

Code Review

This pull request aims to fix out-of-memory (OOM) errors in the model initialization tests, specifically by re-enabling tests that were previously skipped. The changes look solid and well-targeted.

Key changes include:

• Removing the pytest.skip for Glm4vForConditionalGeneration, GraniteSpeechForConditionalGeneration, and KimiVLForConditionalGeneration in tests/models/test_initialization.py.
• Adding an override for audio_config in the test setup to reduce memory usage for audio models, which is a good proactive measure.
• Correcting the initialization of the language model in Glm4vForConditionalGeneration by explicitly passing the text_config. This is a crucial fix that ensures test-specific configuration overrides are properly applied, and it also improves the correctness of the model initialization in general.

The changes are logical and directly address the OOM issue. I don't see any issues of high or critical severity. Well done.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[DarkLight1337]

Thanks, see if the test passes now