Training a 1.93B byte-level LM on one A100: 60K steps, flat to 4x context (huggingface.co) 1 points by EverestAn 1mo ago ↗ HN
1 comment
[ 4.2 ms ] story [ 15.6 ms ] thread