0030a2dd021c5ae0c67536ab64bbe6bf89c6493e,rlkit/torch/her/her.py,HER,_get_action_and_info,#HER#,110

Before Change

        goal = self._rollout_goal
        if self.observation_key:
            observation = observation[self.observation_key]
        if self.desired_goal_key:
            goal = self._rollout_goal[self.desired_goal_key]
        new_obs = np.hstack((observation, goal))
        return self.exploration_policy.get_action(new_obs)

    def get_eval_paths(self):
        paths = []

After Change

        self.exploration_policy.set_num_steps_total(self._n_env_steps_total)
        new_obs = np.hstack((
            observation[self.observation_key],
            observation[self.desired_goal_key],
        ))
        return self.exploration_policy.get_action(new_obs)

Instances