Skip to content

Latest commit

 

History

History
66 lines (42 loc) · 2.03 KB

org_README.md

File metadata and controls

66 lines (42 loc) · 2.03 KB
Raw

NEMATUS

Attention-based encoder-decoder model for neural machine translation

This package is based on the dl4mt-tutorial by Kyunghyun Cho et al. ( https://github.com/nyu-dl/dl4mt-tutorial ). It was used to produce top-scoring systems at the WMT 16 shared translation task.

The changes to Nematus include:

  • ensemble decoding (and new translation API to support it)
  • dropout on all layers (Gal, 2015) http://arxiv.org/abs/1512.05287
  • automatic training set reshuffling between epochs
  • n-best output for decoder
  • more output options (attention weights; word-level probabilities) and visualization scripts
  • performance improvements to decoder
  • rescoring support
  • execute arbitrary validation scripts (for BLEU early stopping)
  • vocabulary files and model parameters are stored in JSON format (backward-compatible loading)

INSTALLATION

Nematus requires the following packages:

  • Python >= 2.7
  • numpy
  • ipdb
  • Theano >= 0.7 (and its dependencies).

we recommend executing the following command in a Python virtual environment: pip install numpy numexpr cython tables theano ipdb

the following packages are optional, but highly recommended

  • CUDA >= 7 (only GPU training is sufficiently fast)
  • cuDNN >= 3 (speeds up training substantially)

you can run Nematus locally. To install it, execute python setup.py install

USAGE INSTRUCTIONS

Change the hard-coded paths to data in test/train_nmt.py then run

THEANO_FLAGS=device=gpu,floatX=float32 python train_nmt.py

To decode, see the command test/test.sh

A sample setup that includes preprocessing is provided at https://github.com/rsennrich/wmt16-scripts/tree/master/sample

PUBLICATIONS

the code is based on the following model:

Dzmitry Bahdanau, Kyunghyun Cho, Yoshua Bengio (2015): Neural Machine Translation by Jointly Learning to Align and Translate, Proceedings of the International Conference on Learning Representations (ICLR).

for the changes specific to Nematus, please consider the following papers:

TBD