fix

ml-agents/mlagents/trainers/ppo_transfer/optimizer.py

                         "value_loss": self.value_loss,
                         "policy_loss": self.abs_policy_loss,
                         "model_loss": self.model_loss,
-                        "reward_loss": self.policy.reward_loss,
+                    }
+                )
+                if self.predict_return:
+                    self.update_dict.update(
+                    {
+                        "reward_loss": self.policy.reward_loss,
                     }
                 )