Bayesian Parameter-Efficient Fine-Tuning for Overcoming Catastrophic Forgetting

TL;DR

This paper introduces a Bayesian approach to parameter-efficient fine-tuning that effectively prevents catastrophic forgetting in models like text-to-speech synthesis and language modeling, using Laplace approximations to regularize the process.

Contribution

It applies Bayesian learning techniques to PEFT with differentiable parameter shifts, demonstrating improved knowledge preservation without performance loss.

Findings

Kronecker-factored approximation better preserves pre-training knowledge

Bayesian regularization prevents catastrophic forgetting in PEFT

No degradation in fine-tuning performance observed

Abstract

We are motivated primarily by the adaptation of text-to-speech synthesis models; however we argue that more generic parameter-efficient fine-tuning (PEFT) is an appropriate framework to do such adaptation. Nevertheless, catastrophic forgetting remains an issue with PEFT, damaging the pre-trained model's inherent capabilities. We demonstrate that existing Bayesian learning techniques can be applied to PEFT to prevent catastrophic forgetting as long as the parameter shift of the fine-tuned layers can be calculated differentiably. In a principled series of experiments on language modeling and speech synthesis tasks, we utilize established Laplace approximations, including diagonal and Kronecker-factored approaches, to regularize PEFT with the low-rank adaptation (LoRA) and compare their performance in pre-training knowledge preservation. Our results demonstrate that catastrophic forgetting can be overcome by our methods without degrading the fine-tuning performance, and using the Kronecker-factored approximation produces a better preservation of the pre-training knowledge than the diagonal ones.