Skip to content

Consider model weight averaging after pretext task? #7

Description

@JoshVarty

I've heard that this can help performance, maybe it helps us here?

Basically, do three runs on the pretext task, take the three sets of weights and average them.

I need to find the paper this is from to make sure I'm understanding it correctly.

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