Dynamic Encoder Size Based on Data-Driven Layer-wise Pruning for Speech Recognition

arXiv (Cornell University)

Abstract

Varying-size models are often required to deploy ASR systems under different hardware and/or application constraints such as memory and latency. To avoid redundant training and optimization efforts for individual models of different sizes, we present the dynamic encoder size approach, which jointly trains multiple performant models within one supernet from scratch. These subnets of various sizes are layer-wise pruned from the supernet, and thus, enjoy full parameter sharing. By combining score-based pruning with supernet training, we propose two novel methods, Simple-Top-k and Iterative-Zero-Out, to automatically select the best-performing subnets in a data-driven manner, avoiding resource-intensive search efforts. Our experiments using CTC on both Librispeech and TED-LIUM-v2 corpora show that our methods can achieve on-par performance as individually trained models of each size category. Also, our approach consistently brings small performance improvements for the full-size supernet.

Authors 5

1. Jingjing Xu Aachen AppTek GmbH Machine Learning and Human Language Technology Group Computer Science Department

   RWTH Aachen University · FH Aachen

   Affiliation as printed

   AppTek GmbH , 52062 Aachen , Germany

   Machine Learning and Human Language Technology Group , Computer Science Dept. , RWTH Aachen University , Germany

2. Wei Guo Zhou Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University ,

3. Zijian Yang Aachen AppTek GmbH Machine Learning and Human Language Technology Group Computer Science Department

   RWTH Aachen University · FH Aachen

   Affiliation as printed

   AppTek GmbH , 52062 Aachen , Germany

   Machine Learning and Human Language Technology Group , Computer Science Dept. , RWTH Aachen University , Germany

4. Eugen Beck Aachen AppTek GmbH

   FH Aachen

   Affiliation as printed

   AppTek GmbH , 52062 Aachen , Germany

5. R. Schlueter