Skip to content

Debug self forcing autograd errors #55

@shahbuland

Description

@shahbuland

Currently the self forcing trainer referenced in configs/dit_v4_sf.yml does not work. Attempting to run it gives weird autograd errors and none of the models parameters receive any gradients. This seems interconnected with a bug that causes illegal memory access errors when attempting cached decoding using flex attention. Further investigation is needed, will log more details when error messages come up.

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