Skip to content

Add Kimi-K2.5 INT4 vLLM v0.16.0 benchmark for MI300X#860

Closed
functionstackx wants to merge 3 commits intomainfrom
claude/issue-859-20260303-0604
Closed

Add Kimi-K2.5 INT4 vLLM v0.16.0 benchmark for MI300X#860
functionstackx wants to merge 3 commits intomainfrom
claude/issue-859-20260303-0604

Conversation

@functionstackx
Copy link
Copy Markdown
Contributor

@functionstackx functionstackx commented Mar 3, 2026

following AMD andy's recipe https://x.com/linluo77/status/2017024513595301985

Add single-node benchmark configuration for Kimi-K2.5 INT4 on MI300X using vLLM v0.16.0, following AMD Andy Luo's recipe. Based on the existing MI355X INT4 Kimi recipe with TP=8, concurrency 4-64.

Closes #859

Generated with Claude Code

Add single-node benchmark configuration for Kimi-K2.5 INT4 on MI300X
using vLLM v0.16.0, following AMD Andy Luo's recipe. Based on the
existing MI355X INT4 Kimi recipe with TP=8, concurrency 4-64.

Closes #859

Co-authored-by: functionstackx <functionstackx@users.noreply.github.com>
@functionstackx
Copy link
Copy Markdown
Contributor Author

#861

Copy link
Copy Markdown
Collaborator

@cquil11 cquil11 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm. let's let sweep pass first

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@functionstackx
Copy link
Copy Markdown
Contributor Author

superceded by #975

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

Development

Successfully merging this pull request may close these issues.

vllm 0.16 single node mi300 kimi k2.5 vllm tp8

2 participants