Asymmetry in Low-Rank Adapters of Foundation Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhu, Jiacheng, Greenewald, Kristjan, Nadjahi, Kimia, Borde, Haitz Sáez de Ocáriz, Gabrielsson, Rickard Brüel, Choshen, Leshem, Ghassemi, Marzyeh, Yurochkin, Mikhail, Solomon, Justin
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916140474695680
author Zhu, Jiacheng
Greenewald, Kristjan
Nadjahi, Kimia
Borde, Haitz Sáez de Ocáriz
Gabrielsson, Rickard Brüel
Choshen, Leshem
Ghassemi, Marzyeh
Yurochkin, Mikhail
Solomon, Justin
author_facet Zhu, Jiacheng
Greenewald, Kristjan
Nadjahi, Kimia
Borde, Haitz Sáez de Ocáriz
Gabrielsson, Rickard Brüel
Choshen, Leshem
Ghassemi, Marzyeh
Yurochkin, Mikhail
Solomon, Justin
contents Parameter-efficient fine-tuning optimizes large, pre-trained foundation models by updating a subset of parameters; in this class, Low-Rank Adaptation (LoRA) is particularly effective. Inspired by an effort to investigate the different roles of LoRA matrices during fine-tuning, this paper characterizes and leverages unexpected asymmetry in the importance of low-rank adapter matrices. Specifically, when updating the parameter matrices of a neural network by adding a product $BA$, we observe that the $B$ and $A$ matrices have distinct functions: $A$ extracts features from the input, while $B$ uses these features to create the desired output. Based on this observation, we demonstrate that fine-tuning $B$ is inherently more effective than fine-tuning $A$, and that a random untrained $A$ should perform nearly as well as a fine-tuned one. Using an information-theoretic lens, we also bound the generalization of low-rank adapters, showing that the parameter savings of exclusively training $B$ improves the bound. We support our conclusions with experiments on RoBERTa, BART-Large, LLaMA-2, and ViTs.
format Preprint
id arxiv_https___arxiv_org_abs_2402_16842
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Asymmetry in Low-Rank Adapters of Foundation Models
Zhu, Jiacheng
Greenewald, Kristjan
Nadjahi, Kimia
Borde, Haitz Sáez de Ocáriz
Gabrielsson, Rickard Brüel
Choshen, Leshem
Ghassemi, Marzyeh
Yurochkin, Mikhail
Solomon, Justin
Machine Learning
Parameter-efficient fine-tuning optimizes large, pre-trained foundation models by updating a subset of parameters; in this class, Low-Rank Adaptation (LoRA) is particularly effective. Inspired by an effort to investigate the different roles of LoRA matrices during fine-tuning, this paper characterizes and leverages unexpected asymmetry in the importance of low-rank adapter matrices. Specifically, when updating the parameter matrices of a neural network by adding a product $BA$, we observe that the $B$ and $A$ matrices have distinct functions: $A$ extracts features from the input, while $B$ uses these features to create the desired output. Based on this observation, we demonstrate that fine-tuning $B$ is inherently more effective than fine-tuning $A$, and that a random untrained $A$ should perform nearly as well as a fine-tuned one. Using an information-theoretic lens, we also bound the generalization of low-rank adapters, showing that the parameter savings of exclusively training $B$ improves the bound. We support our conclusions with experiments on RoBERTa, BART-Large, LLaMA-2, and ViTs.
title Asymmetry in Low-Rank Adapters of Foundation Models
topic Machine Learning
url https://arxiv.org/abs/2402.16842