[Bugfix] fix cuda 12.6 and 11.8 build

[jinzhen-lin]

fix cuda 12.6 and 11.8 build introduced by #22428

[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.

🚀

[gemini-code-assist]

Code Review

This pull request aims to fix CUDA build issues for versions 11.8 and 12.6 by conditionally applying the -static-global-template-stub=false compile flag. The change introduces a version check before setting the flag. However, there's a potential inconsistency: the code checks for CUDA version >= 12.8, while the pull request title mentions fixing version 12.6. This could lead to the flag being incorrectly disabled for CUDA versions 12.6 and 12.7. I've added comments suggesting to align the version check with the intended fix, likely by changing 12.8 to 12.6.

[jeejeelee]

@mgoin Could you plz look at this PR?

[mgoin]

Ahh, sorry for making you use this flag 😓