MultiPath/NA-NMT

Issues

• universal transformer

  #2 opened 5 years ago by 520jefferson
  1

• how to get the dataset ？

  #3 opened 5 years ago by wangwang110
  0

• Shouldn't the loss be normalized by the batch_size, not the number of words in a batch?

  #1 opened 7 years ago by hyhieu
  1