OpenNMT - Python

An implementation of Seq2Seq Neural Machine Translation that actually works. I want to make it easy for people to train their own seq2seq model with any corpus. I am also adding the parameters of my trained model for people to just use it without training. If you have a model that works share your model params here, as external link or do a pull request. I have used Cornell Movie Dialog Corpus to train my model. A link to preprocessed data and scripts for preprocessing can be found in this repo.

Have Fun!

Setup

• Create temporary working directory prior to training

mkdir working_dir

• Download test/train data from Cornell Movie Dialog Corpus

cd data/
bash pull_data.sh

Training

# edit seq2seq.ini file to set 
#		mode = train
python execute.py
# or use custom ini file
#		python execute.py my_custom_conf.ini

Testing

# edit seq2seq.ini file to set 
#		mode = test
python execute.py