Learning Deep Transformer Models for Machine Translation

TL;DR

This paper demonstrates that a properly designed deep Transformer model with advanced normalization and layer passing techniques can outperform standard Transformer-Big models in machine translation, while being smaller and faster to train.

Contribution

The paper introduces a novel deep Transformer architecture that surpasses Transformer-Big in translation quality, size, and training speed by using improved normalization and layer passing methods.

Findings

Deep Transformer models outperform Transformer-Big in BLEU scores.

Deep models are 1.6 times smaller and 3 times faster to train.

Proper normalization and layer passing are key to training very deep Transformers.

Abstract

Transformer is the state-of-the-art model in recent machine translation evaluations. Two strands of research are promising to improve models of this kind: the first uses wide networks (a.k.a. Transformer-Big) and has been the de facto standard for the development of the Transformer system, and the other uses deeper language representation but faces the difficulty arising from learning deep networks. Here, we continue the line of research on the latter. We claim that a truly deep Transformer model can surpass the Transformer-Big counterpart by 1) proper use of layer normalization and 2) a novel way of passing the combination of previous layers to the next. On WMT'16 English- German, NIST OpenMT'12 Chinese-English and larger WMT'18 Chinese-English tasks, our deep system (30/25-layer encoder) outperforms the shallow Transformer-Big/Base baseline (6-layer encoder) by 0.4-2.4 BLEU points. As another bonus, the deep model is 1.6X smaller in size and 3X faster in training than Transformer-Big.