Implementation of ReWiND, "Language-Guided Rewards Teach Robot Policies without New Demonstrations", from USC / Amazon Robotics - View it on GitHub