cc1019054695/bon-nat

Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation

PyTorch implementation of the models described in the paper Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation .

Dependencies

Python

• Python 3.6
• PyTorch >= 0.4
• Numpy
• NLTK
• torchtext 0.2.1
• torchvision
• revtok
• multiset
• ipdb

Related code

• This code is based on dl4mt-nonauto and RSI-NAT. We mainly modified the model.py (line 1107-1292).

Downloading Datasets

The original translation corpora can be downloaded from (IWLST'16 En-De, WMT'16 En-Ro, WMT'14 En-De). We recommend you to download the preprocessed corpora released in dl4mt-nonauto. Set correct path to data in data_path() function located in data.py before you run the code.

BoN-Joint

Combine the BoN objective and the cross-entropy loss to train NAT from scratch. This process usually takes about 5 days.

$ sh joint_wmt.sh

Take a checkpoint and train the length prediction model. This process usually takes about 1 day.

$ sh tune_wmt.sh

Decode the test set. This process usually takes about 20 seconds.

$ sh decode_wmt.sh

BoN-FT

First, train a NAT model using the cross-entropy loss. This process usually takes about 5 days.

$ sh mle_wmt.sh

Then, take a pre-trained checkpoint and finetune the NAT model using the BoN objective. This process usually takes about 3 hours.

$ sh bontune_wmt.sh

Take a finetuned checkpoint and train the length prediction model. This process usually takes about 1 day.

$ sh tune_wmt.sh

Decode the test set. This process usually takes about 20 seconds.

$ sh decode_wmt.sh

Reinforce-NAT

We also implement Reinforce-NAT (line 1294-1390) described in the paper Retrieving Sequential Information for Non-Autoregressive Neural Machine Translation. See RSI-NAT for the usage.

Citation

If you find the resources in this repository useful, please consider citing:

@article{Shao:19,
  author    = {Chenze Shao, Jinchao Zhang, Yang Feng, Fandong Meng and Jie Zhou},
  title     = {Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation},
  year      = {2019},
  journal   = {arXiv preprint arXiv:1911.09320},
}