pytorch-trpo(Hessian-vector product version)

This is a PyTorch implementation of "Trust Region Policy Optimization (TRPO)" with exact Hessian-vector product instead of finite differences approximation.