Comment by sgu999

For those also wondering: https://pile.eleuther.ai

> The Pile is a 825 GiB diverse, open source language modelling data set that consists of 22 smaller, high-quality datasets combined together.

By what's the legal complication with it?