Hugging Face dataset · #805

Zyphra/Zyda

Zyphra/Zyda is #805 among listed Hugging Face datasets, with 84 likes. Zyda is a 1.3T language modeling dataset created by collecting open and high quality datasets and combining them and performing a uniform filtering and deduplication step. We find that Zyda performs extremely well in ablations and is at least comparable and potentially better…