- pytorch=1.6
- transformers=3.1.0
- spacy=3.0 # conflicts with transformers
- RAMS (Download at [https://nlp.jhu.edu/rams/], and put it at data)
- Run
python preprocess.py
for preprocessing, which will generate adata.pk
file (We have released the back translation results in the fileback_translation_result.txt
to ease the preprocessing steps). - Run
train.py
for model training, which will save the model at themodels
directory. - Run
evaluate.py
for model evaluation.