Increasing Model Capacity for Free: A Simple Strategy for Parameter Efficient Fine-tuning

TL;DR

CapaBoost is a straightforward method that enhances the capacity of parameter-efficient fine-tuning by using low-rank updates and random masking, leading to improved performance across various tasks without extra costs.

Contribution

It introduces CapaBoost, a novel strategy that increases model capacity via low-rank updates with random masks, compatible with existing fine-tuning methods.

Findings

Significant performance improvements on NLP and image tasks.

No additional computational or storage costs.

Effective across diverse downstream applications.

Abstract

Fine-tuning large pre-trained foundation models, such as the 175B GPT-3, has attracted more attention for downstream tasks recently. While parameter-efficient fine-tuning methods have been proposed and proven effective without retraining all model parameters, their performance is limited by the capacity of incremental modules, especially under constrained parameter budgets. \\ To overcome this challenge, we propose CapaBoost, a simple yet effective strategy that enhances model capacity by leveraging low-rank updates through parallel weight modules in target layers. By applying static random masks to the shared weight matrix, CapaBoost constructs a diverse set of weight matrices, effectively increasing the rank of incremental weights without adding parameters. Notably, our approach can be seamlessly integrated into various existing parameter-efficient fine-tuning methods. We extensively validate the efficacy of CapaBoost through experiments on diverse downstream tasks, including natural language understanding, question answering, and image classification. Our results demonstrate significant improvements over baselines, without incurring additional computation or storage costs. Our code is available at \url{https://github.com/LINs-lab/CapaBoost}.