Improve normalizer for scalar

async_worker/continuous_actor_critic.py

@@ -48,7 +48,7 @@ class ContinuousAdvantageActorCritic:
 
             steps += 1
             total_reward += reward
-            reward = np.asscalar(self.reward_normalizer(np.array([reward])))
+            reward = self.reward_normalizer(reward)
 
             if deterministic:
                 if terminal:

async_worker/ppo.py

@@ -79,7 +79,7 @@ class ProximalPolicyOptimization:
                 batched_steps += 1
                 episode_length += 1
 
-                reward = np.asscalar(self.reward_normalizer(np.array([reward])))
+                reward = self.reward_normalizer(reward)
                 rewards.append(reward)
 
                 if done: