Skip to content

Conserms about the reproduction of pi*0.6 #925

@qianweijiujiu

Description

@qianweijiujiu

Hi, I confront an issue during my production of pi0.6
Which is how the negative samples work.
Opensource reproduction of the pi
0.6 trains the negative but I wonder why we have to learn these negative samples?
Do we really need that. It's kind of confusing and I can not understand.
Would u plz add some ablation study experients in the appendix of your paper or you official website.

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