train_policy_gradient_tf.py 文件源码

def tf_discount_rewards(self, tf_r):  # tf_r ~ [game_steps,1]
        discount_f = lambda a, v: a * self._gamma + v;
        tf_r_reverse = tf.scan(discount_f, tf.reverse(tf_r, [True, False]))
        tf_discounted_r = tf.reverse(tf_r_reverse, [True, False])
        return tf_discounted_r