Video Pre Training Versions Save

Hey all! We just wanted to let you know we have shared two things:

1. the inverse dynamics model for predicting players' actions from unlabeled video, and
2. an example behavioural cloning code to further train the models