Skip to content

about dataset proccess #21

@DiaQusNet

Description

@DiaQusNet

In the function sharegpt_format, it seems that both the user's questions and the responses are concatenated together into a single {"text": text} format dataset. Why is the response included in the dataset? According to the documentation for the trl, it supports direct input in a conversational format.

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