We show that variational learning can significantly improve the accuracy and calibration of Low-Rank Adaptation (LoRA) without a substantial increase in the cost. We replace AdamW by the Improved Variational Online Newton (IVON) algorithm to finetune large language models. For Llama-2 with 7 billion parameters, IVON improves the accuracy over AdamW by 2.8% and expected calibration error by 4.6%. The accuracy is also better than the other Bayesian alternatives, yet the cost is lower and the implementation is easier. Our work provides additional evidence for the effectiveness of IVON for large language models. The code is available at https://github.com/team-approx-bayes/ivon-lora.
@article{arxiv.2411.04421,
title = {Variational Low-Rank Adaptation Using IVON},
author = {Bai Cong and Nico Daheim and Yuesong Shen and Daniel Cremers and Rio Yokota and Mohammad Emtiyaz Khan and Thomas Möllenhoff},
journal= {arXiv preprint arXiv:2411.04421},
year = {2024}
}
Comments
Published at 38th Workshop on Fine-Tuning in Machine Learning (NeurIPS 2024). Code available at https://github.com/team-approx-bayes/ivon-lora. In version 2 we fixed a typo in the equation of prior in section 2