Dynamic Multi-Branch Layers for On-Device Neural Machine Translation

TL;DR

This paper introduces a dynamic multi-branch layer approach for on-device neural machine translation, significantly improving translation quality and speed while maintaining low resource consumption.

Contribution

It proposes a layer-wise dynamic multi-branch network with shared-private reparameterization, enabling efficient on-device NMT with enhanced performance.

Findings

Up to 1.7 BLEU improvement on WMT14 En-De

Up to 1.8 BLEU improvement on WMT20 Zh-En

Up to 1.5x faster inference with same parameters

Abstract

With the rapid development of artificial intelligence (AI), there is a trend in moving AI applications, such as neural machine translation (NMT), from cloud to mobile devices. Constrained by limited hardware resources and battery, the performance of on-device NMT systems is far from satisfactory. Inspired by conditional computation, we propose to improve the performance of on-device NMT systems with dynamic multi-branch layers. Specifically, we design a layer-wise dynamic multi-branch network with only one branch activated during training and inference. As not all branches are activated during training, we propose shared-private reparameterization to ensure sufficient training for each branch. At almost the same computational cost, our method achieves improvements of up to 1.7 BLEU points on the WMT14 English-German translation task and 1.8 BLEU points on the WMT20 Chinese-English translation task over the Transformer model, respectively. Compared with a strong baseline that also uses multiple branches, the proposed method is up to 1.5 times faster with the same number of parameters.