Skip to content

Reproduce efficiency results #12

@Yunge6666

Description

@Yunge6666

I can't reproduce the FPS results in the paper. It seems to me that the current SlideAttention still generates a lot of intermediates, which limits bandwidth and memory. Please correct me if I'm wrong.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions