An Efficient Split Fine-tuning Framework for Edge and Cloud Collaborative Learning

TL;DR

This paper introduces an efficient split fine-tuning framework for edge and cloud collaborative learning, significantly reducing communication costs while maintaining model accuracy across multiple NLP datasets.

Contribution

It presents a novel framework with matrix decomposition-based compression, link elimination, and PyTorch integration for efficient edge-cloud model fine-tuning.

Findings

Reduces communication traffic by 96 times

Maintains high model accuracy across 9 NLP datasets

Enables easy extension of existing training scripts

Abstract

To enable the pre-trained models to be fine-tuned with local data on edge devices without sharing data with the cloud, we design an efficient split fine-tuning (SFT) framework for edge and cloud collaborative learning. We propose three novel techniques in this framework. First, we propose a matrix decomposition-based method to compress the intermediate output of a neural network to reduce the communication volume between the edge device and the cloud server. Second, we eliminate particular links in the model without affecting the convergence performance in fine-tuning. Third, we implement our system atop PyTorch to allow users to easily extend their existing training scripts to enjoy the efficient edge and cloud collaborative learning. Experiments results on 9 NLP datasets show that our framework can reduce the communication traffic by 96 times with little impact on the model accuracy.