arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2312.03415cs.LG
收藏

Run LoRA Run:更快、更轻量的 LoRA 实现

Run LoRA Run: Faster and Lighter LoRA Implementations

Skolkovo Institute of Science and Technology(斯科尔科沃科技学院);Artificial Intelligence Research Institute(人工智能研究院)

Daria Cherniuk, Aleksandr Mikhalev, Ivan Oseledets

AI总结:

本文提出 RunLoRA 框架,依据线性层维度、输入维度和秩,并结合 FLOPs 与时间估计选择前向、反向计算图,以在不损失准确率的前提下加速 LoRA 训练与微调,最高提速 28%。

摘要:

LoRA 是一种通过在线性层中引入低秩适配器来减少神经网络可训练参数数量的技术。该技术既用于大语言模型的微调,也用于其完整训练。本文提出 RunLoRA 框架,用于高效实现 LoRA,显著提升使用低秩适配器进行神经网络训练和微调的速度。所提出的实现根据相应线性层的维度、层输入维度以及 lora rank 来优化 LoRA 运算的计算,基于 FLOPs 和时间估计选择最佳的前向与反向计算图,从而在不牺牲准确率的情况下加快训练。实验结果表明,在语言建模网络上最高可实现 28% 的加速。

英文摘要:

LoRA is a technique that reduces the number of trainable parameters in a neural network by introducing low-rank adapters to linear layers. This technique is used both for fine-tuning and full training of large language models. This paper presents the RunLoRA framework for efficient implementations of LoRA that significantly improves the speed of neural network training and fine-tuning using low-rank adapters. The proposed implementation optimizes the computation of LoRA operations based on dimensions of corresponding linear layer, layer input dimensions and lora rank by choosing best forward and backward computation graph based on FLOPs and time estimations, resulting in faster training without sacrificing accuracy. The experimental results show up to 28\% speedup on language modeling networks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑