Skip to content

why activated param is 1.8B? #2

@LinB203

Description

@LinB203

The total model is 3B, right? And you have one text expert and one image expert. In every model forward, the input will be split two set, the text tokens are fed into text expert and image tokens are fed into image expert. So, it seems that all param are calculated in model forward. If I miss anything, please correct me. Thanks

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