Adversarially Trained Weighted Actor-Critic for Safe Offline Reinforcement Learning

Additionally, we offer a practical version of WSAC and compare it with existing state-of-the-art safe offline RL algorithms in several continuous control environments.