Fix demo loader and remaining tests

ml-agents/mlagents/trainers/demo_loader.py

     sequence_length: int,
 ) -> AgentBuffer:
     # Create and populate buffer using experiences
-    demo_buffer = AgentProcessorBuffer()
-    update_buffer = AgentBuffer()
+    demo_process_buffer = AgentProcessorBuffer()
+    demo_buffer = AgentBuffer()
     for idx, experience in enumerate(pair_infos):
         if idx > len(pair_infos) - 2:
             break
         previous_action = np.array(pair_infos[idx].action_info.vector_actions) * 0
         if idx > 0:
             previous_action = np.array(pair_infos[idx - 1].action_info.vector_actions)
-        demo_buffer[0].last_brain_info = current_brain_info
-        demo_buffer[0]["done"].append(next_brain_info.local_done[0])
-        demo_buffer[0]["rewards"].append(next_brain_info.rewards[0])
+        demo_process_buffer[0].last_brain_info = current_brain_info
+        demo_process_buffer[0]["done"].append(next_brain_info.local_done[0])
+        demo_process_buffer[0]["rewards"].append(next_brain_info.rewards[0])
-            demo_buffer[0]["visual_obs%d" % i].append(
+            demo_process_buffer[0]["visual_obs%d" % i].append(
-            demo_buffer[0]["vector_obs"].append(
+            demo_process_buffer[0]["vector_obs"].append(
-        demo_buffer[0]["actions"].append(current_pair_info.action_info.vector_actions)
-        demo_buffer[0]["prev_action"].append(previous_action)
+        demo_process_buffer[0]["actions"].append(
+            current_pair_info.action_info.vector_actions
+        )
+        demo_process_buffer[0]["prev_action"].append(previous_action)
-            demo_buffer.append_update_buffer(
-                update_buffer, 0, batch_size=None, training_length=sequence_length
+            demo_process_buffer.append_update_buffer(
+                demo_buffer, 0, batch_size=None, training_length=sequence_length
-            demo_buffer.reset_local_buffers()
-    demo_buffer.append_update_buffer(
-        update_buffer, 0, batch_size=None, training_length=sequence_length
+            demo_process_buffer.reset_local_buffers()
+    demo_process_buffer.append_update_buffer(
+        demo_buffer, 0, batch_size=None, training_length=sequence_length
     )
     return demo_buffer
 

ml-agents/mlagents/trainers/tests/test_demo_loader.py

     assert len(pair_infos) == total_expected
 
     _, demo_buffer = demo_to_buffer(path_prefix + "/test.demo", 1)
-    assert len(demo_buffer.update_buffer["actions"]) == total_expected - 1
+    assert len(demo_buffer["actions"]) == total_expected - 1
 
 
 def test_load_demo_dir():
     assert len(pair_infos) == total_expected
 
     _, demo_buffer = demo_to_buffer(path_prefix + "/test_demo_dir", 1)
-    assert len(demo_buffer.update_buffer["actions"]) == total_expected - 1
+    assert len(demo_buffer["actions"]) == total_expected - 1

ml-agents/mlagents/trainers/tests/test_reward_signals.py

 def reward_signal_update(env, policy, reward_signal_name):
     buffer = mb.simulate_rollout(env, policy, BUFFER_INIT_SAMPLES)
     feed_dict = policy.reward_signals[reward_signal_name].prepare_update(
-        policy.model, buffer.update_buffer.make_mini_batch(0, 10), 2
+        policy.model, buffer.make_mini_batch(0, 10), 2
     )
     out = policy._execute_model(
         feed_dict, policy.reward_signals[reward_signal_name].update_dict