8000 Model architecture consultation · Issue #20 · aehrc/cxrmate · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
Model architecture consultation #20
Open
@yihp

Description

@yihp

hi,
I would like to ask why the visual encoder was chosen as CvT and why the decoder in this paper uses a 6-layer Transformer. What is the basis for these choices? Did you refer to other works or conduct any comparative experiments?

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

      0