Skip to content

Support Timestep level caching? #840

Description

@eitanturok

Hi there, love the project!

Are there plans to support timestep-level caching techniques such as TeaCache, MagCache, TaoCache, EasyCache, Chipmunk, and more? These methods allow for skipping entire diffusion timesteps, which would complement the existing block-level caching beautifully.

There is debate about supporting timestep-level caching directly in sglang sgl-project/sglang#18498.

Thanks!

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions