Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning

Fine-tuning large language models is computationally expensive, and while existing parameter-efficient methods like Low-Rank Adaptation (LoRA) reduce computational costs, they are limited by suboptimal initialization strategies. We introduce Activation-Guided LoRA (AG-LoRA), a novel approach that in...

Full description

Saved in:
Bibliographic Details
Main Authors: Qingchen Wang, Shengyu Shen
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10852296/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850141358464958464
author Qingchen Wang
Shengyu Shen
author_facet Qingchen Wang
Shengyu Shen
author_sort Qingchen Wang
collection DOAJ
description Fine-tuning large language models is computationally expensive, and while existing parameter-efficient methods like Low-Rank Adaptation (LoRA) reduce computational costs, they are limited by suboptimal initialization strategies. We introduce Activation-Guided LoRA (AG-LoRA), a novel approach that initializes LoRA modules using Singular Value Decomposition (SVD) guided by activation patterns. Our method employs pre-trained weights combined with activation-based weighting factors and implements a new global rank assignment strategy that accounts for activation outliers. Experimental evaluations on LLaMA and CLIP models show that AG-LoRA achieves superior performance while reducing GPU memory usage compared to existing methods. In tests with LLaMA 7B models, AG-LoRA reached 75.9% accuracy across various tasks, surpassing both LoRA and DoRA baselines. AG-LoRA demonstrates significant improvements in parameter-efficient fine-tuning of large language models, offering enhanced performance and reduced computational requirements. These advances make it a promising solution for efficient model adaptation across diverse applications.
format Article
id doaj-art-e5927d68407e4329b0854c0fac2bf388
institution OA Journals
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-e5927d68407e4329b0854c0fac2bf3882025-08-20T02:29:27ZengIEEEIEEE Access2169-35362025-01-0113709097091810.1109/ACCESS.2025.353370110852296Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-TuningQingchen Wang0https://orcid.org/0009-0002-9175-1944Shengyu Shen1https://orcid.org/0009-0009-8079-5176Zijin Research and Innovation Center, Nanjing Yunwen Network Technology Company Ltd., Nanjing, Jiangsu, ChinaZijin Research and Innovation Center, Nanjing Yunwen Network Technology Company Ltd., Nanjing, Jiangsu, ChinaFine-tuning large language models is computationally expensive, and while existing parameter-efficient methods like Low-Rank Adaptation (LoRA) reduce computational costs, they are limited by suboptimal initialization strategies. We introduce Activation-Guided LoRA (AG-LoRA), a novel approach that initializes LoRA modules using Singular Value Decomposition (SVD) guided by activation patterns. Our method employs pre-trained weights combined with activation-based weighting factors and implements a new global rank assignment strategy that accounts for activation outliers. Experimental evaluations on LLaMA and CLIP models show that AG-LoRA achieves superior performance while reducing GPU memory usage compared to existing methods. In tests with LLaMA 7B models, AG-LoRA reached 75.9% accuracy across various tasks, surpassing both LoRA and DoRA baselines. AG-LoRA demonstrates significant improvements in parameter-efficient fine-tuning of large language models, offering enhanced performance and reduced computational requirements. These advances make it a promising solution for efficient model adaptation across diverse applications.https://ieeexplore.ieee.org/document/10852296/Deep learninglow-rank adaptationparameter-efficient fine-tuning
spellingShingle Qingchen Wang
Shengyu Shen
Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning
IEEE Access
Deep learning
low-rank adaptation
parameter-efficient fine-tuning
title Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning
title_full Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning
title_fullStr Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning
title_full_unstemmed Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning
title_short Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning
title_sort activation guided low rank parameter adaptation for efficient model fine tuning
topic Deep learning
low-rank adaptation
parameter-efficient fine-tuning
url https://ieeexplore.ieee.org/document/10852296/
work_keys_str_mv AT qingchenwang activationguidedlowrankparameteradaptationforefficientmodelfinetuning
AT shengyushen activationguidedlowrankparameteradaptationforefficientmodelfinetuning