RingAda: Pipelining Large Model Fine-Tuning on Edge Devices with Scheduled Layer Unfreezing

TL;DR

RingAda is a novel framework enabling efficient, privacy-preserving fine-tuning of large transformer models on edge devices through scheduled layer unfreezing and collaborative pipeline training.

Contribution

It introduces a ring topology and pipeline mechanism for parameter-efficient, accelerated fine-tuning of large models on edge devices.

Findings

Reduces fine-tuning time significantly

Lowers memory costs during training

Maintains competitive model performance

Abstract

To enable large model (LM) based edge intelligent service provisioning, on-device fine-tuning with locally personalized data allows for continuous and privacy-preserving LM customization. In this paper, we propose RingAda, a collaborative training framework designed for fine-tuning transformer-based LMs on edge devices. Particularly, RingAda performs parameter-efficient adapter fine-tuning across a set of interconnected edge devices, forming a ring topology for per-batch training by sequentially placing frozen transformer blocks and their trainable adapter modules on the devices. RingAda follows a novel pipeline-parallel training mechanism with top-down adapter unfreezing, allowing for early-stopping of backpropagation at the lowest unfrozen adapter layer, thereby accelerating the fine-tuning process. Extensive experimental results demonstrate that RingAda significantly reduces fine-tuning time and memory costs while maintaining competitive model performance compared to its peer designs.