-

ml-agents/mlagents/trainers/simple_env_manager.py

         self.previous_all_action_info = all_action_info
 
         for brain_name, action_info in all_action_info.items():
-            self.env.set_actions(brain_name, np.array(action_info.action))
+            try:
+                self.env.set_actions(brain_name, np.array(action_info.action))
+            except:
+                pass
         self.env.step()
         all_step_result = self._generate_all_results()