563acaf48cf475f4b125fd32374c6bb566049551,src/garage/torch/algos/vpg.py,VPG,process_samples,#VPG#Any#Any#,268

Before Change

        for path in paths:
            path["returns"] = tensor_utils.discount_cumsum(
                path["rewards"], self.discount)
        average_discounted_return = (np.mean(
            [path["returns"][0] for path in paths]))
        undiscounted_returns = [sum(path["rewards"]) for path in paths]
        average_return = np.mean(undiscounted_returns)

        self._episode_reward_mean.extend(undiscounted_returns)

After Change

                                            total_length=self.max_path_length,
                                            axis=0) for path in paths
        ])
        rewards = torch.stack([
            loss_function_utils.pad_to_last(path["rewards"],
                                            total_length=self.max_path_length)
            for path in paths
        ])

        return valids, obs, actions, rewards

    def _log(self, itr, paths):

Instances