Fix pcl_rl bugs (#10465)

Author: dkqkxx

research/pcl_rl/controller.py

@@ -204,8 +204,8 @@ def sample_episodes(self, sess, greedy=False):
        observations, actions, rewards,
        pads) = self._sample_episodes(sess, greedy=greedy)
 
-      observations = zip(*observations)
-      actions = zip(*actions)
+      observations = list(zip(*observations))
+      actions = list(zip(*actions))
 
       terminated = np.array(self.env.dones)
 

research/pcl_rl/env_spec.py

@@ -175,7 +175,7 @@ def initial_obs(self, batch_size):
     if batched:
       return obs
     else:
-      return zip(*obs)[0]
+      return list(zip(*obs))[0]
 
   def initial_act(self, batch_size=None):
     batched = batch_size is not None
@@ -191,7 +191,7 @@ def initial_act(self, batch_size=None):
     if batched:
       return act
     else:
-      return zip(*act)[0]
+      return list(zip(*act))[0]
 
   def is_discrete(self, typ):
     return typ == spaces.discrete