News for hackers. Stuff that matters.

A comment on The Pile: a dataset for language modeling [pdf]. Parent comment

Re:The Pile: a dataset for language modeling [pdf] by moffkalast on Wednesday July 12, 2023 @04:33AM (#36691829)
> They DMCA’ed one of my repos distributing LLaMA

Boy they'll be mad once they learn about Huggingface distributing thousands of LLama fine tunes with full weights.

[ Parent | on HN ]
Re:The Pile: a dataset for language modeling [pdf] by redox99 on Wednesday July 12, 2023 @05:16AM (#36692096)
Weights being copyrightable is already a questionable thing. Derivatives (like finetunes) is even more questionable.
[ Parent | on HN ]