Python package for neural machine translation built with Keras and Tensorflow
nmt is library that was developed as part of diploma thesis. It can train LSTM encoder-decoder model based on provided datasets. Even though machine translation is its primary use, it can be used for any other similar training, such as to train chat bots. It only depends on the provided datasets from which the model learns how to response to given sequence. Published as paper on Excel@FIT 2018 conference.
One of those methods can be used:
- Install package globally from wheel in dist folder (
pip install dist/nmt.whl) - Install package globally with python
setup.py install - Put it in the project folder
Use it with import nmt
Complete documentation of Translator api is here
Exmaple of script running nmt library main.py
- Prebuilt documentation is in docs/_build/
- Can be build from source with
sphinx-build -b html ./docs ./docs/_build - Live version from github
- Test are written using pytest
- Can be run with
pytest tests
- better decomposition
- upload to PyPi