Compacter: Efficient Low-Rank Hypercomplex Adapter Layers

TL;DR

Compacter is a parameter-efficient fine-tuning method for large language models that uses hypercomplex layers and Kronecker products to achieve performance comparable to full fine-tuning while training only a tiny fraction of parameters.

Contribution

It introduces a novel hypercomplex adapter layer that significantly reduces trainable parameters while maintaining or improving task performance.

Findings

Compacter trains only 0.047% of parameters and matches or exceeds standard fine-tuning performance.

It outperforms existing parameter-efficient methods on GLUE and SuperGLUE benchmarks.

The method is effective in low-resource settings.

Abstract

Adapting large-scale pretrained language models to downstream tasks via fine-tuning is the standard method for achieving state-of-the-art performance on NLP benchmarks. However, fine-tuning all weights of models with millions or billions of parameters is sample-inefficient, unstable in low-resource settings, and wasteful as it requires storing a separate copy of the model for each task. Recent work has developed parameter-efficient fine-tuning methods, but these approaches either still require a relatively large number of parameters or underperform standard fine-tuning. In this work, we propose Compacter, a method for fine-tuning large-scale language models with a better trade-off between task performance and the number of trainable parameters than prior work. Compacter accomplishes this by building on top of ideas from adapters, low-rank optimization, and parameterized hypercomplex multiplication layers. Specifically, Compacter inserts task-specific weight matrices into a pretrained model's weights, which are computed efficiently as a sum of Kronecker products between shared "slow" weights and "fast" rank-one matrices defined per Compacter layer. By only training 0.047% of a pretrained model's parameters, Compacter performs on par with standard fine-tuning on GLUE and outperforms standard fine-tuning on SuperGLUE and low-resource settings. Our code is publicly available at~\url{https://github.com/rabeehk/compacter}.