Hacker News new | ask | show | jobs
by whimsicalism 1166 days ago
Right but the way that this paper proposes determining the best rank is by training a LoRA with the full rank.