AlphaLoRA: Assigning LoRA Experts Based on Layer Training Quality

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Qing, Peijun, Gao, Chongyang, Zhou, Yefan, Diao, Xingjian, Yang, Yaoqing, Vosoughi, Soroush
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910648728813568
author Qing, Peijun
Gao, Chongyang
Zhou, Yefan
Diao, Xingjian
Yang, Yaoqing
Vosoughi, Soroush
author_facet Qing, Peijun
Gao, Chongyang
Zhou, Yefan
Diao, Xingjian
Yang, Yaoqing
Vosoughi, Soroush
contents Parameter-efficient fine-tuning methods, such as Low-Rank Adaptation (LoRA), are known to enhance training efficiency in Large Language Models (LLMs). Due to the limited parameters of LoRA, recent studies seek to combine LoRA with Mixture-of-Experts (MoE) to boost performance across various tasks. However, inspired by the observed redundancy in traditional MoE structures, previous studies identify similar redundancy among LoRA experts within the MoE architecture, highlighting the necessity for non-uniform allocation of LoRA experts across different layers. In this paper, we leverage Heavy-Tailed Self-Regularization (HT-SR) Theory to design a fine-grained allocation strategy. Our analysis reveals that the number of experts per layer correlates with layer training quality, which exhibits significant variability across layers. Based on this, we introduce AlphaLoRA, a theoretically principled and training-free method for allocating LoRA experts to further mitigate redundancy. Experiments on three models across ten language processing and reasoning benchmarks demonstrate that AlphaLoRA achieves comparable or superior performance over all baselines. Our code is available at https://github.com/morelife2017/alphalora.
format Preprint
id arxiv_https___arxiv_org_abs_2410_10054
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle AlphaLoRA: Assigning LoRA Experts Based on Layer Training Quality
Qing, Peijun
Gao, Chongyang
Zhou, Yefan
Diao, Xingjian
Yang, Yaoqing
Vosoughi, Soroush
Computation and Language
Parameter-efficient fine-tuning methods, such as Low-Rank Adaptation (LoRA), are known to enhance training efficiency in Large Language Models (LLMs). Due to the limited parameters of LoRA, recent studies seek to combine LoRA with Mixture-of-Experts (MoE) to boost performance across various tasks. However, inspired by the observed redundancy in traditional MoE structures, previous studies identify similar redundancy among LoRA experts within the MoE architecture, highlighting the necessity for non-uniform allocation of LoRA experts across different layers. In this paper, we leverage Heavy-Tailed Self-Regularization (HT-SR) Theory to design a fine-grained allocation strategy. Our analysis reveals that the number of experts per layer correlates with layer training quality, which exhibits significant variability across layers. Based on this, we introduce AlphaLoRA, a theoretically principled and training-free method for allocating LoRA experts to further mitigate redundancy. Experiments on three models across ten language processing and reasoning benchmarks demonstrate that AlphaLoRA achieves comparable or superior performance over all baselines. Our code is available at https://github.com/morelife2017/alphalora.
title AlphaLoRA: Assigning LoRA Experts Based on Layer Training Quality
topic Computation and Language
url https://arxiv.org/abs/2410.10054