isabella232 / nonauto-nmt

PyTorch Implementation of "Non-Autoregressive Neural Machine Translation"

Non-Autoregressive Transformer

Code release for Non-Autoregressive Neural Machine Translation by Jiatao Gu, James Bradbury, Caiming Xiong, Victor O.K. Li, and Richard Socher.

Requires PyTorch 0.3, torchtext 0.2.1, and SpaCy.

The pipeline for training a NAT model for a given language pair includes:

1. run_alignment_wmt_LANG.sh (runs fast_align for alignment supervision)
2. run_LANG.sh (trains an autoregressive model)
3. run_LANG_decode.sh (produces the distillation corpus for training the NAT)
4. run_LANG_fast.sh (trains the NAT model)
5. run_LANG_fine.sh (fine-tunes the NAT model)