Install the requirements using
pip install -r requirements.txt
After that run the run.sh or python main.py for the training and the fix_nonenglish.py is a post-process script to map the character back to the original non-english one.
Model output on WebNLG 2020 test set is available at output.txt.