Code cleanup

async_worker/actor_critic.py

@@ -29,7 +29,6 @@ class AdvantageActorCritic:
             prob, log_prob, value = self.worker_network.predict(np.stack([state]))
             action = self.policy.sample(prob.data.numpy().flatten(), deterministic)
             next_state, reward, terminal, _ = self.task.step(action)
-            terminal = (terminal or (self.config.max_episode_length and steps > self.config.max_episode_length))
 
             steps += 1
             total_reward += reward

async_worker/continuous_actor_critic.py

@@ -43,7 +43,6 @@ class ContinuousAdvantageActorCritic:
                                         False)
             action = self.config.action_shift_fn(action)
             next_state, reward, terminal, _ = self.task.step(action)
-            terminal = (terminal or (config.max_episode_length and steps >= config.max_episode_length))
             next_state = self.state_normalizer(next_state)
 
             steps += 1

async_worker/dpg.py

@@ -58,7 +58,6 @@ class DeterministicPolicyGradient:
             if not deterministic:
                 action += self.random_process.sample()
             next_state, reward, done, info = self.task.step(action)
-            done = (done or (config.max_episode_length and steps >= config.max_episode_length))
             next_state = self.state_normalizer(next_state)
             total_reward += reward
             reward = self.reward_normalizer(reward)

async_worker/n_step_q.py

@@ -30,7 +30,6 @@ class NStepQLearning:
             q = self.worker_network.predict(np.stack([state]))
             action = self.policy.sample(q.data.numpy().flatten(), deterministic)
             next_state, reward, terminal, _ = self.task.step(action)
-            terminal = (terminal or (config.max_episode_length and steps >= config.max_episode_length))
 
             steps += 1
             total_reward += reward

async_worker/one_step_q.py

@@ -30,7 +30,6 @@ class OneStepQLearning:
             q = self.worker_network.predict(np.stack([state]))
             action = self.policy.sample(q.data.numpy().flatten(), deterministic)
             next_state, reward, terminal, _ = self.task.step(action)
-            terminal = (terminal or (config.max_episode_length and steps >= config.max_episode_length))
 
             steps += 1
             total_reward += reward

async_worker/one_step_sarsa.py

@@ -30,7 +30,6 @@ class OneStepSarsa:
         pending = []
         while not config.stop_signal.value:
             next_state, reward, terminal, _ = self.task.step(action)
-            terminal = (terminal or (config.max_episode_length and steps >= config.max_episode_length))
             next_q = self.worker_network.predict(np.stack([next_state]))
             next_action = self.policy.sample(next_q.data.numpy().flatten(), deterministic)
             pending.append([q, action, reward, next_state, next_action])

async_worker/ppo.py

@@ -72,7 +72,6 @@ class ProximalPolicyOptimization:
                 values.append(value)
                 state, reward, done, _ = self.task.step(action)
                 state = self.state_normalizer(state)
-                done = (done or (config.max_episode_length and episode_length > config.max_episode_length))
 
                 batched_rewards += reward
                 batched_steps += 1