Etra: a machine translation framework based on attentional neural network techniques
Requirements
Minimum requirements:
- Python 2.7.x
- Numpy
- Chainer
Installation
Install with CPUs
Beware that using CPUs for training is very slow, we strongly recommend to install etra on machines which have GPU cards.
pip install -e .
Install with GPUs
etra
is built on top of chainer
, so to install etra
on GPU machines, make sure that you have CUDA
installed correctly, follow
the CUDA instruction.
Then, run
pip install -e .
After installation complete, you can verify whether etra
has been installed with GPU correctly by running,
python -c "import chainer; print chainer.cuda.available"
If output is "False", you have NOT configure CUDA properly.
Examples
Examples on how to build an end-to-end MT system for any language pairs can be found at egs