Skip to content

Loading pretrained EVA-02 weights using RoPE at different resolution #167

@CardenB

Description

@CardenB

Hi,

Does anyone have any guidance on how to load pretrained weights using RoPE for different resolution input image for fine tuning?

It's unclear to me how EVA-02 went for 1024 MIM checkpoints fine tuned at 1536 for o365. I'd personally like to do the inverse.

What are the primary issues to avoid? Is it the changing size for the global attention indexes?

Appreciate any insight people have to offer. Let me know if I can share more information to make the discussion easier.

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