Read news on Zyda with our app.
Read more in the app
Zyphra debuts Zyda, a 1.3T language modeling dataset it claims outperforms Pile, C4, arxiv