English

LoRA+: Efficient Low Rank Adaptation of Large Models

Machine Learning 2024-07-08 v2 Artificial Intelligence Computation and Language Machine Learning

Abstract

In this paper, we show that Low Rank Adaptation (LoRA) as originally introduced in Hu et al. (2021) leads to suboptimal finetuning of models with large width (embedding dimension). This is due to the fact that adapter matrices A and B in LoRA are updated with the same learning rate. Using scaling arguments for large width networks, we demonstrate that using the same learning rate for A and B does not allow efficient feature learning. We then show that this suboptimality of LoRA can be corrected simply by setting different learning rates for the LoRA adapter matrices A and B with a well-chosen ratio. We call this proposed algorithm LoRA++. In our extensive experiments, LoRA++ improves performance (1-2 %\% improvements) and finetuning speed (up to \sim 2X SpeedUp), at the same computational cost as LoRA.

Keywords

Cite

@article{arxiv.2402.12354,
  title  = {LoRA+: Efficient Low Rank Adaptation of Large Models},
  author = {Soufiane Hayou and Nikhil Ghosh and Bin Yu},
  journal= {arXiv preprint arXiv:2402.12354},
  year   = {2024}
}

Comments

27 pages

R2 v1 2026-06-28T14:53:29.248Z