3 ms·The Pile: An 800GB Dataset of Diverse Text for Language Modeling [pdf]1 points by nixtaken 6y agodang 6y agohttps://news.ycombinator.com/item?id=25607809 https://news.ycombinator.com/item?id=25607809