Skip to content
View original post on X: vLLMOfficial· 41/100AI score41/100

vLLM releases an open-source agent skill for debugging model diffs

AISummary

vLLM says it distilled its workflow for finding and fixing numerical divergence into an open-source agent skill, pull request 61551 on GitHub.

The workflow finds the first diverging op, builds a minimal repro, fixes it, re-verifies, then profiles and optimizes.

The post says DeepSeek-V4.1 has reached zero diff, with a follow-up coming, and that end-to-end training wall-clock speedup remains to be measured.

Post on XView on X
vLLMVerified on X
@vllm_project

Part of a thread · earlier post

5/ We distilled the workflow into an open-source agent skill:

Find the first diverging op → minimal repro → fix → re-verify → profile and optimize.

Skill: https://github.com/vllm-project/vllm/pull/61551
Recipes and code are linked in the blog. DeepSeek-V4.1 has also reached 0 diff; a follow-up is coming.

Thanks to Yan Bai (NVIDIA DevTech) and Ao Shen (Inferact)!

Scope: each baseline used one configuration with R3, without TIS/MIS, and was not thoroughly tuned. End-to-end training wall-clock speedup remains to be measured.

Source: vLLM · x.comPublished · added here

Topics

NVIDIAAI agentsOpen-source ecosystemDeployment & engineeringTutorials & how-to