9e9c4fd3322b6e8f47572fefdb8fd65018fb96f7,rl_coach/exploration_policies/e_greedy.py,EGreedy,get_action,#EGreedy#Any#,85

Before Change

        // step the epsilon schedule and generate a new random value for next time
        if self.phase == RunPhase.TRAIN:
            self.epsilon_schedule.step()
        self.current_random_value = np.random.rand()
        return chosen_action

    def get_control_param(self):

After Change

            if self.current_random_value < epsilon:
                chosen_action = self.action_space.sample()
                probabilities = np.full(len(self.action_space.actions),
                                      1. / (self.action_space.high[0] - self.action_space.low[0] + 1))
            else:
                chosen_action = np.argmax(action_values)

Instances