Skip to content

Efficient training on OOS datasetsΒ #74

@George3d6

Description

@George3d6

I'm curious if there are any examples of fine-tuning the model, either using the transformers version or, ideally the galai library directly. I assume it's not too different from any other model but I'd still feel more confident doing so with some examples.

I'd also be curious to know if there are examples or attempts at training this model specifically to answer questions based on a person's understanding, i.e. either:

  1. Using a dataset where the outputs are the model's own outputs adjusted to be understandable by that person or
  2. Using a dataset of writing and conversation from that person to hopefully pick-up on speech patterns that make sense to them.

I've been toying around with this idea using scibert and a t5 trained on a similar corpus but this model seem superior to both.

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