Structured Pruning of Self-Supervised Pre-trained Models for Speech Recognition and Understanding

TL;DR

This paper introduces three structured pruning methods tailored for heterogeneous speech models combining convolutional frontends and Transformer backends, achieving significant computational reduction with maintained or improved accuracy.

Contribution

It presents novel task-specific structured pruning techniques for speech models with both convolutional and Transformer components, addressing a gap in prior Transformer-focused pruning studies.

Findings

Achieves 10-30% less computation with better accuracy than original models.

Reduces computation by 40-50% without accuracy loss.

Demonstrates effectiveness on LibriSpeech and SLURP datasets.

Abstract

Self-supervised speech representation learning (SSL) has shown to be effective in various downstream tasks, but SSL models are usually large and slow. Model compression techniques such as pruning aim to reduce the model size and computation without degradation in accuracy. Prior studies focus on the pruning of Transformers; however, speech models not only utilize a stack of Transformer blocks, but also combine a frontend network based on multiple convolutional layers for low-level feature representation learning. This frontend has a small size but a heavy computational cost. In this work, we propose three task-specific structured pruning methods to deal with such heterogeneous networks. Experiments on LibriSpeech and SLURP show that the proposed method is more accurate than the original wav2vec2-base with 10% to 30% less computation, and is able to reduce the computation by 40% to 50% without any degradation.