Federated Fine-Tuning of Sparsely-Activated Large Language Models on Resource-Constrained Devices

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Fahao, Wan, Jie, Li, Peng, Su, Zhou, Yu, Dongxiao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917000723300352
author Chen, Fahao
Wan, Jie
Li, Peng
Su, Zhou
Yu, Dongxiao
author_facet Chen, Fahao
Wan, Jie
Li, Peng
Su, Zhou
Yu, Dongxiao
contents Federated fine-tuning of Mixture-of-Experts (MoE)-based large language models (LLMs) is challenging due to their massive computational requirements and the resource constraints of participants. Existing working attempts to fill this gap through model quantization, computation offloading, or expert pruning. However, they cannot achieve desired performance due to impractical system assumptions and a lack of consideration for MoE-specific characteristics. In this paper, we propose FLUX, a system designed to enable federated fine-tuning of MoE-based LLMs across participants with constrained computing resources (e.g., consumer-grade GPUs), aiming to minimize time-to-accuracy. FLUX introduces three key innovations: (1) quantization-based local profiling to estimate expert activation with minimal overhead, (2) adaptive layer-aware expert merging to reduce resource consumption while preserving accuracy, and (3) dynamic expert role assignment using an exploration-exploitation strategy to balance tuning and non-tuning experts. Extensive experiments on LLaMA-MoE and DeepSeek-MoE with multiple benchmark datasets demonstrate that FLUX significantly outperforms existing methods, achieving up to 4.75X speedup in time-to-accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2508_19078
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Federated Fine-Tuning of Sparsely-Activated Large Language Models on Resource-Constrained Devices
Chen, Fahao
Wan, Jie
Li, Peng
Su, Zhou
Yu, Dongxiao
Distributed, Parallel, and Cluster Computing
Artificial Intelligence
Federated fine-tuning of Mixture-of-Experts (MoE)-based large language models (LLMs) is challenging due to their massive computational requirements and the resource constraints of participants. Existing working attempts to fill this gap through model quantization, computation offloading, or expert pruning. However, they cannot achieve desired performance due to impractical system assumptions and a lack of consideration for MoE-specific characteristics. In this paper, we propose FLUX, a system designed to enable federated fine-tuning of MoE-based LLMs across participants with constrained computing resources (e.g., consumer-grade GPUs), aiming to minimize time-to-accuracy. FLUX introduces three key innovations: (1) quantization-based local profiling to estimate expert activation with minimal overhead, (2) adaptive layer-aware expert merging to reduce resource consumption while preserving accuracy, and (3) dynamic expert role assignment using an exploration-exploitation strategy to balance tuning and non-tuning experts. Extensive experiments on LLaMA-MoE and DeepSeek-MoE with multiple benchmark datasets demonstrate that FLUX significantly outperforms existing methods, achieving up to 4.75X speedup in time-to-accuracy.
title Federated Fine-Tuning of Sparsely-Activated Large Language Models on Resource-Constrained Devices
topic Distributed, Parallel, and Cluster Computing
Artificial Intelligence
url https://arxiv.org/abs/2508.19078