Distillation-based Layer Dropping (DLD): Effective End-to-end Framework for Dynamic Speech Networks
This addresses the need for efficient, adaptive speech recognition models on edge devices, offering an incremental improvement over existing layer dropping methods.
The paper tackles the performance degradation in dynamic speech networks when using layer dropping for computational efficiency, proposing a distillation-based framework that reduces word error rate by 9.32% and 2.25% for high and no dropping cases with 33.3% faster training.
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.