opt-peter-1.3B / latest

Commit History

add new checkpoint trained for a hundred steps with smaller max grad norm and weight decay
7a20e92

pszemraj commited on

update model with approx 1.6 epochs training
687ed54

pszemraj commited on