@inproceedings{sharpmodelsondullhardwarefastand,
  title = {Sharp Models on Dull Hardware: Fast and Accurate Neural Machine Translation Decoding on the CPU},
  author = {Jacob Devlin},
  year = {2017},
  booktitle = {EMNLP 2017 9},
  eprint = {1705.01991},
  archivePrefix = {arXiv},
  url = {http://arxiv.org/abs/1705.01991v1},
}