Fork of https://github.com/wgrathwohl/BackpropThroughTheVoidRL with modifications for the paper "The Mirage of Action-Dependent Baselines in Reinforcement Learning". - View it on GitHub