Personalized Collaborative Fine-Tuning for On-Device Large Language Models

Nicolas Wagner, Dongyang Fan, Martin Jaggi
4/15/2024
cs.CLcs.LG

No official code found. Best implementation: epfml/personalized-collaborative-llms (6 stars, Apache-2.0, last commit 2024-09-02).

Abstract

We explore on-device self-supervised collaborative fine-tuning of large language models with limited local data availability. Taking inspiration from the collaborative learning community, we introduce three distinct trust-weighted gradient aggregation schemes: weight similarity-based, prediction similarity-based and validation performance-based. To minimize communication overhead, we integrate Low-Rank Adaptation (LoRA) and only exchange LoRA weight updates. Our protocols, driven by prediction and performance metrics, surpass both FedAvg and local fine-tuning methods, which is particularly evident in realistic scenarios with more diverse local data distributions. The results underscore the effectiveness of our approach in addressing heterogeneity and scarcity within local datasets.

View on arXivView PDF

Code Implementations(1)

epfml/personalized-collaborative-llmsCommunity100%
Apache-2.0

Exploration on-device self-supervised collaborative fine-tuning of large language models with limited local data availability, using Low-Rank Adaptation (LoRA). We introduce three distinct trust-weighted gradient aggregation schemes: weight similarity-based, prediction similarity-based and validation performance-based.

60Shell, PythonApr 8, 20242 years agoApache-2.0

Similar Papers

Resource-Efficient Federated Fine-Tuning Large Language Models for Heterogeneous Data

85% match2025

HSplitLoRA: A Heterogeneous Split Parameter-Efficient Fine-Tuning Framework for Large Language Models

85% match2025

Personalized Federated Fine-tuning for Heterogeneous Data: An Automatic Rank Learning Approach via Two-Level LoRA

85% match2025

FLoRIST: Singular Value Thresholding for Efficient and Accurate Federated Fine-Tuning of Large Language Models

85% match1 repo2025

Decentralized Low-Rank Fine-Tuning of Large Language Models

84% match2025

Exploring Gradient Subspaces: Addressing and Overcoming LoRA's Limitations in Federated Fine-Tuning of Large Language Models

84% match2024

Cite this paper

@article{wagner2024personalized,
  title  = {Personalized Collaborative Fine-Tuning for On-Device Large Language Models},
  author = {Nicolas Wagner and Dongyang Fan and Martin Jaggi},
  year   = {2024},
  eprint = {2404.09753},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2404.09753v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2404.09753.svg)](https://codeofpaper.com/paper/2404.09753)

Discussion