vllm
[Perf] Use value-only reduction for native per-token FP8 quantization
#59800

Merged

[Perf] Use value-only reduction for native per-token FP8 quantization #59800

mgoin merged 2 commits into vllm-project:main from jackLei0901:perf/native-per-token-fp8-amax

[Perf] Use value-only reduction for native per-token FP8 quantization

ad887521

jackLei0901 requested a review from

mgoin 16 days ago

jackLei0901 requested a review from

robertgshaw2-redhat 16 days ago

jackLei0901 requested a review from

tlrmchlsmth 16 days ago

jackLei0901 requested a review from

yewentao256 16 days ago

jackLei0901 requested a review from

pavanimajety 16 days ago

jackLei0901 requested a review from

zyongye 16 days ago

claude commented on 2026-10-02

mergify added quantization

mgoin approved these changes on 2026-10-02

mgoin added verified

Merge branch 'main' into perf/native-per-token-fp8-amax

13ae6763

mgoin enabled auto-merge (squash) 16 days ago

github-actions added ready

mgoin merged 10f6ab01 into main 16 days ago

Reviewers

mgoin

claude

robertgshaw2-redhat

tlrmchlsmth

yewentao256

pavanimajety

zyongye

Assignees

No one assigned

Labels

ready quantization verified

Milestone

No milestone