Distillation-based Layer Dropping (DLD): Effective End-to-end Framework for Dynamic Speech Networks

TL;DR

This paper introduces a distillation-based layer dropping framework for dynamic speech networks, improving performance and efficiency on speech recognition tasks by effectively combining knowledge distillation with layer dropping.

Contribution

The proposed DLD framework effectively integrates knowledge distillation with layer dropping, achieving state-of-the-art results in dynamic speech recognition models.

Findings

Reduces word error rate by 9.32% in high dropping scenarios.

Achieves 33.3% reduction in training time.

Improves performance for both high and no dropping cases.

Abstract

Edge devices operate in constrained and varying resource settings, requiring dynamic architectures that can adapt to limitations of the available resources. To meet such demands, layer dropping ($\mathcal{LD}$) approach is typically used to transform static models into dynamic ones by skipping parts of the network along with reducing overall computational complexity. However, existing $\mathcal{LD}$ methods greatly impact the dynamic model's performance for low and high dropping cases, deteriorating the performance-computation trade-off. To this end, we propose a distillation-based layer dropping (DLD) framework that effectively combines the capabilities of knowledge distillation and $\mathcal{LD}$ in an end-to-end fashion, thereby achieving state-of-the-art performance for dynamic speech networks. Comprehensive experimentation utilizing well-known speech recognition methods, including conformer and WavLM, on three public benchmarks demonstrates the effectiveness of our framework, reducing the word error rate by $9.32\%$ and $2.25\%$ for high and no dropping cases with $33.3\%$ reduction in training time.