vllm
[torch.compile] Use FakeTensors instead of real GPU tensors for single-size compilation
#36093

Merged

[torch.compile] Use FakeTensors instead of real GPU tensors for single-size compilation #36093

zou3519 merged 1 commit into vllm-project:main from zou3519:compile_size_fake

zou3519 requested a review from

youkaichao 7 days ago

zou3519 requested a review from

ProExpertProg 7 days ago

zou3519 requested a review from

BoyuanFeng 7 days ago

gemini-code-assist commented on 2026-03-05

zou3519 force pushed from 6f5635a9 to bcb9e5ba 7 days ago

mergify added needs-rebase

BoyuanFeng approved these changes on 2026-03-09

zou3519 force pushed from bcb9e5ba to d19ac4a6 3 days ago

mergify removed needs-rebase

zou3519 added ready

zou3519 enabled auto-merge (squash) 2 days ago

Use FakeTensors instead of real GPU tensors for single-size compilation

7c46bdb4

zou3519 force pushed from 17d977c6 to 7c46bdb4 1 day ago

zou3519 merged 822e250a into main 1 day ago

Reviewers

BoyuanFeng

gemini-code-assist

youkaichao

ProExpertProg

Assignees

No one assigned

Labels

ready

Milestone

No milestone