4 ms·The Pile: An 800GB Dataset of Diverse Text for Language Modeling2 points by andyxor 5y agoandyxor 5y agohttps://pile.eleuther.ai/ https://pile.eleuther.ai/