arXiv 2603.044362026-02-19cs.LGcs.AI

ZorBA: 零阶联邦微调LLM的异构块激活

ZorBA: Zeroth-order Federated Fine-tuning of LLMs with Heterogeneous Block Activation

Department of Electrical and Computer Engineering, The University of British Columbia, Vancouver, Canada（电气与计算机工程系，不列颠哥伦比亚大学）；Department of Computer Science and Engineering, Southern University of Science and Technology, China（计算机科学与工程系，南方科技大学）

Chuiyang Meng, Ming Tang, Vincent W. S. Wong

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

ZorBA通过异构块激活和零阶优化，有效降低LLM联邦微调的VRAM使用和通信开销。

摘要：

联邦微调大型语言模型（LLMs）能够实现分布式客户端之间的协同调优。然而，由于LLMs的规模较大，联邦学习（FL）中的本地更新可能导致显著的视频随机访问内存（VRAM）使用。此外，频繁的模型交换可能导致显著的通信开销。为解决这些问题，本文提出了ZorBA，一种基于零阶优化的联邦微调框架，具有异构块激活机制。ZorBA利用零阶优化在客户端上消除梯度存储，通过前向传递实现。ZorBA包含异构块激活机制，其中中央服务器将不同的transformer块子集分配给客户端，以加速收敛速率并减少VRAM使用。此外，ZorBA利用共享随机种子和梯度的有限差分来减少通信开销。我们进行了理论分析，以表征块激活决策对收敛速率和VRAM使用的影响。为了共同提高收敛速率并减少VRAM使用，我们提出了一个优化问题来优化块激活决策。我们提出了一种$ε$-约束词典序算法来解决这个问题。实验结果表明，ZorBA在VRAM使用上比三个联邦微调基线高出62.41%，并且通信开销较低。

英文摘要：

Federated fine-tuning of large language models (LLMs) enables collaborative tuning across distributed clients. However, due to the large size of LLMs, local updates in federated learning (FL) may incur substantial video random-access memory (VRAM) usage. Moreover, frequent model exchange may lead to significant communication overhead. To tackle these challenges, in this paper we propose ZorBA, a zeroth-order optimization-based federated fine-tuning framework with heterogeneous block activation. ZorBA leverages zeroth-order optimization to eliminate the storage of gradients at the clients by forward passes. ZorBA includes a heterogeneous block activation mechanism in which the central server allocates different subsets of transformer blocks to clients in order to accelerate the convergence rate and reduce the VRAM usage. Furthermore, ZorBA utilizes shared random seeds and the finite differences of gradients in order to reduce the communication overhead. We conduct theoretical analysis to characterize the effect of block activation decisions on the convergence rate and VRAM usage. To jointly enhance the convergence rate and reduce the VRAM usage, we formulate an optimization problem to optimize the block activation decisions. We propose an $ε$-constraint lexicographic algorithm to solve this problem. Experimental results show that ZorBA outperforms three federated fine-tuning baselines in VRAM usage by up to 62.41% and incurs a low communication overhead.

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。