KEMBAR78
Make auto_functionalize HOPs cacheable · Issue #151188 · pytorch/pytorch · GitHub
Skip to content

Make auto_functionalize HOPs cacheable #151188

@zou3519

Description

@zou3519

I think this should go into 2.7.1. This was the reason that sglang had torch.compile caching issues and the fix is very simple.

cc @chauhang @penguinwu @ydwu4 @bdhirsh

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions