← Back to storiesTraining a 1.93B byte-level LM on one A100: 60K steps, flat to 4x context1 pointsby EverestAn 2 hours ago0 commentsDiscussionLoading discussion