Efficient Inference For Neural Machine Translation
This work addresses inference efficiency for machine translation users, but it is incremental as it optimizes existing methods rather than introducing new ones.
The paper tackled the problem of slow inference in neural machine translation by combining known techniques like simplified recurrent units and attention pruning, achieving up to 109% speedup on CPU and 84% on GPU with a 25% parameter reduction while maintaining BLEU scores.
Large Transformer models have achieved state-of-the-art results in neural machine translation and have become standard in the field. In this work, we look for the optimal combination of known techniques to optimize inference speed without sacrificing translation quality. We conduct an empirical study that stacks various approaches and demonstrates that combination of replacing decoder self-attention with simplified recurrent units, adopting a deep encoder and a shallow decoder architecture and multi-head attention pruning can achieve up to 109% and 84% speedup on CPU and GPU respectively and reduce the number of parameters by 25% while maintaining the same translation quality in terms of BLEU.