vllm
[torch.compile] Turn on silu+fp4 quant fusion by default for O1+
#34718

Merged

[torch.compile] Turn on silu+fp4 quant fusion by default for O1+ #34718

ProExpertProg merged 2 commits into vllm-project:main from neuralmagic:luka/silu-quant-o1-fp4

Turn on silu+fp4 quant fusion by default

238f271d

ProExpertProg requested a review from

WoosukKwon 165 days ago

ProExpertProg requested a review from

youkaichao 165 days ago

ProExpertProg requested a review from

robertgshaw2-redhat 165 days ago

ProExpertProg requested a review from

mgoin 165 days ago

ProExpertProg requested a review from

tlrmchlsmth 165 days ago

ProExpertProg requested a review from

houseroad 165 days ago

ProExpertProg requested a review from

hmellor 165 days ago

ProExpertProg requested a review from

yewentao256 165 days ago

gemini-code-assist commented on 2026-02-17

mgoin approved these changes on 2026-02-17

mgoin added performance

mgoin added ready

mgoin added torch.compile

mgoin added nvidia

Fix model_config=None

2e558592

ProExpertProg enabled auto-merge (squash) 165 days ago

ProExpertProg merged 02e8f26c into main 165 days ago

Reviewers

mgoin

gemini-code-assist

WoosukKwon

youkaichao

robertgshaw2-redhat

tlrmchlsmth

houseroad

hmellor

yewentao256

Assignees

No one assigned

Labels

performance ready torch.compile nvidia

Milestone

No milestone

vllm [torch.compile] Turn on silu+fp4 quant fusion by default for O1+ #34718 Merged

[torch.compile] Turn on silu+fp4 quant fusion by default for O1+ #34718

vllm
[torch.compile] Turn on silu+fp4 quant fusion by default for O1+
#34718

Merged