[Bugfix] Fix cutlass dispatch for fp8/int8 to properly invoke M<=16 c…

[Ther-LF]

Fixed a bug where cutlass dispatch for fp8 and int8 couldn't invoke the M<=16 config.

Changes

• Fixed the condition check for fp8/int8 dispatch to properly handle M<=16 cases
• Ensured the correct kernel config is selected for small matrix sizes

[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.

🚀

[Ther-LF]

I tested the performance of meta-llama/Llama-2-7b-hf-TP1 with a token length of 16, comparing FP8 and INT8 precision in Cutlass W8A8 GEMM. The original performance was


, and after optimizations, it improved to


. The results show a consistent speedup of 10-20%.

[Ther-LF]

Hi @mgoin ,

Would you mind checking my PR and merging it if possible? You previously reviewed a related PR [Kernel] Tuned int8 kernels for Ada Lovelace

Thanks for your time!

[mgoin]

Seems reasonable to me thanks for the results. cc @varun-sundar-rabindranath