Empirical Evaluation of Deep Learning Model Compression Techniques on the WaveNet Vocoder
arXiv:2011.10469
Abstract
WaveNet is a state-of-the-art text-to-speech vocoder that remains challenging to deploy due to its autoregressive loop. In this work we focus on ways to accelerate the original WaveNet architecture directly, as opposed to modifying the architecture, such that the model can be deployed as part of a scalable text-to-speech system. We survey a wide variety of model compression techniques that are amenable to deployment on a range of hardware platforms. In particular, we compare different model sparsity methods and levels, and seven widely used precisions as targets for quantization; and are able to achieve models with a compression ratio of up to 13.84 without loss in audio fidelity compared to a dense, single-precision floating-point baseline. All techniques are implemented using existing open source deep learning frameworks and libraries to encourage their wider adoption.
References in corpus (8)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
- Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation
- Block-Sparse Recurrent Neural Networks
- FastWave: Accelerating Autoregressive Convolutional Neural Networks on FPGA
- QPyTorch: A Low-Precision Arithmetic Simulation Framework
- FeatherWave: An efficient high-fidelity neural vocoder with multi-band linear prediction
- Computation Error Analysis of Block Floating Point Arithmetic Oriented Convolution Neural Network Accelerator Design