Shallow-to-Deep Training for Neural Machine Translation
This addresses efficiency issues in training deep NMT models for machine translation practitioners, but it is incremental as it builds on existing Transformer architectures.
The paper tackles the problem of training deep encoders for neural machine translation being time-consuming by proposing a shallow-to-deep training method that stacks shallow models, resulting in a 54-layer encoder that trains 1.4 times faster than from scratch and achieves BLEU scores of 30.33 and 43.29 on WMT tasks.
Deep encoders have been proven to be effective in improving neural machine translation (NMT) systems, but training an extremely deep encoder is time consuming. Moreover, why deep models help NMT is an open question. In this paper, we investigate the behavior of a well-tuned deep Transformer system. We find that stacking layers is helpful in improving the representation ability of NMT models and adjacent layers perform similarly. This inspires us to develop a shallow-to-deep training method that learns deep models by stacking shallow models. In this way, we successfully train a Transformer system with a 54-layer encoder. Experimental results on WMT'16 English-German and WMT'14 English-French translation tasks show that it is $1.4$ $\times$ faster than training from scratch, and achieves a BLEU score of $30.33$ and $43.29$ on two tasks. The code is publicly available at https://github.com/libeineu/SDT-Training/.