# for pretraining bert
bunkai~=1.4.0
pytextspan>=0.5.4
tensorflow>=2.5.0
tensorflow_datasets>=4.3.0
