Skip to content

How many models does the rl training process have to train? #3230

@aijianiula0601

Description

@aijianiula0601

I saw that the rm and sft models in the source code were put into the triton service. So which models should be loaded to train the rl training process?

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions