This repository is an unofficial PyTorch implementation of ICLR 2019 paper IMPROVING SEQUENCE-TO-SEQUENCE LEARNING VIA OPTIMAL TRANSPORT
Official TensorFlow implementation is here.
This repository uses Weights and Biases.
docker build -t ot-seq2seq-pytorch docker/
docker run --rm -it --gpus device=0 -v $PWD:/work -w /work ot-seq2seq-pytorch bash
or simply install requirements:
python -m pip install -r docker/requirements.txt