35a7a5b27a93a2442f82628abd18d47200901f3d,torch_ac/algos/ppo.py,PPOAlgo,update_parameters,#PPOAlgo#,20

Before Change

                // Compute loss

                rdist, value = self.acmodel.get_rdist_n_value(Variable(b.obs))

                log_dist = F.log_softmax(rdist, dim=1)
                dist = F.softmax(rdist, dim=1)

After Change

                // Compute loss

                rdist = self.acmodel.get_rdist(Variable(b.obs))
                value = self.acmodel.get_value(Variable(b.obs))

                log_dist = F.log_softmax(rdist, dim=1)
                dist = F.softmax(rdist, dim=1)
                entropy = -(log_dist * dist).sum(dim=1).mean()

Instances