Skip to content

Question about the optimizer #85

@pspdada

Description

@pspdada

Hello! I have some questions about the optimizer used in your training. The paper mentions, "We use an Adam optimizer [47] for all the experiments." However, I noticed that the code uses the AdamW optimizer instead, as stated in the training configuration .yaml file with the line optim: adamw_torch. Could you please clarify this?

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