fix: Extend compute capabilities during vllm installation

[dmitry-tokarev-nv]

Overview:

Details:

Where should the reviewer start?

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• closes GitHub issue: #xxx

Summary by CodeRabbit

• New Features
  • Added support for additional CUDA architectures, enabling compatibility with a broader range of GPU hardware configurations.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Walkthrough

The TORCH_CUDA_ARCH_LIST variable in the vLLM installation script was expanded to include additional CUDA architectures (10.3, 12.0, 12.1) alongside the existing 9.0 and 10.0 entries, broadening GPU compatibility for prebuilt wheel and kernel selection during installation.

Changes

Cohort / File(s)	Summary
vLLM CUDA Architecture Configuration container/deps/vllm/install_vllm.sh	Expanded TORCH_CUDA_ARCH_LIST to support additional CUDA compute architectures (10.3, 12.0, 12.1) in addition to previously supported versions (9.0, 10.0)

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Poem

🐰 With whiskers twitching in delight,
More CUDA arches, shining bright!
From ten to twelve, the list does grow,
GPU gardens start to glow! 🚀✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The description contains only template placeholders with no actual content filled in; all sections are empty or contain placeholder text.	Provide detailed explanations in Overview and Details sections, specify which files to review, and link the actual GitHub issue number.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: extending CUDA compute capabilities in the vllm installation script.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

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

---

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.}