8000 Why is drafter trained over all scales? · Issue #6 · czg1225/CoDe · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
Why is drafter trained over all scales? #6
New issue
New issue
Open
Open
Why is drafter trained over all scales?#6
@wjc2830

Description

@wjc2830
wjc2830
opened on Feb 6, 2025

As claimed in Section 3.2, one of motivations is to reduce the interference between small and large scales training dependencies. However, in your practical implementation, both drafter and refiner are trained over all scales, especially there is no special design for drafter training. Can you explain on this?

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
      0