EPIC TTS Models: Empirical Pruning Investigations Characterizing Text-To-Speech Models

TL;DR

This paper investigates the effects of different sparsity techniques on TTS models, finding that pruning during training can be efficient and effective, with neuron removal being more detrimental than parameter removal.

Contribution

It is the first study to compare various sparsity paradigms in text-to-speech synthesis, providing insights into their impact on performance and efficiency.

Findings

Pruning during training matches post-training pruning performance.

Neuron removal degrades quality more than parameter removal.

Sparse models can be trained faster with comparable naturalness and intelligibility.

Abstract

Neural models are known to be over-parameterized, and recent work has shown that sparse text-to-speech (TTS) models can outperform dense models. Although a plethora of sparse methods has been proposed for other domains, such methods have rarely been applied in TTS. In this work, we seek to answer the question: what are the characteristics of selected sparse techniques on the performance and model complexity? We compare a Tacotron2 baseline and the results of applying five techniques. We then evaluate the performance via the factors of naturalness, intelligibility and prosody, while reporting model size and training time. Complementary to prior research, we find that pruning before or during training can achieve similar performance to pruning after training and can be trained much faster, while removing entire neurons degrades performance much more than removing parameters. To our best knowledge, this is the first work that compares sparsity paradigms in text-to-speech synthesis.