finetuning support for pytorch version

[huyz1117]

Will you release the finetuning code for pytorch version?

[siriuz42]

Not on our roadmap at this moment. That said you'd need a data loader and a training loop and that should be it.

The data loader needs to be customized to work on your own datasets, and that's one of the reasons we don't have off-the-shelf finetuning code (besides the main reason that we're short of bandwidth 😮‍💨 ).