Skip to content

Domino + PP #950

@XZQshiyu

Description

@XZQshiyu

I’m excited about the recent introduction of Domino and its impressive TP optimization.
When I was using deepspeed-domino to better overlap comm & comp in TP, I found domino use forward_backward_no_pipelining() in schedules.py. Is that mean I couldn't use domino(tp optimization) and pp together?

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