merge add fire

config/ppo/Tennis.yaml

 behaviors:
   Tennis:
+    framework: pytorch
     trainer_type: ppo
     hyperparameters:
       batch_size: 2048

ml-agents/mlagents/trainers/policy/torch_policy.py

         return self.get_current_step()
 
     def load_weights(self, values: List[np.ndarray]) -> None:
-        pass
+        self.actor_critic.load_state_dict(values)
 
     def init_load_weights(self) -> None:
         pass