Skip to content

Investigate observed fp8 vs fp16 input projections performance difference #35

Open
Open
Investigate observed fp8 vs fp16 input projections performance difference#35

Description

@garykbrixi
garykbrixi
opened on Jan 30, 2025

40b, 1b models have significantly lower performance when use_fp8_input_projections is set to false

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

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions