vllm-project/vllm-ascend

[Misc]: Discussion on accuracy variance

Open

#6,623 opened on Feb 9, 2026

View on GitHub
 (1 comment) (0 reactions) (0 assignees)C++ (1,318 forks)github user discovery
help wantedwait-feedback

Repository metrics

Stars
 (2,180 stars)
PR merge metrics
 (Avg merge 5d 16h) (419 merged PRs in 30d)

Description

Anything you want to discuss about vllm on ascend.

Because of batch variance, We cannot guarantee that the same input will yield the same output in a multi-batch inference case. And this accuracy variance is more explicit in certain datasets.

model dataset acc acc variance
deepseek-v3.1 GPQA 74~82 8
qwen3-235b GPQA 64~71 7
qwen3-480b GPQA 60~67 7

We need to test these case on GPU or use other inference engine, such as sglang to check if this is an acc bug.

If GPU also has the same acc variance like us, we believe that this acc variance is reasonable for this dataset. Otherwise, we need to solve this acc bug.

Contributor guide