A trained model still needs processing resources when it answers a question. That is why inference infrastructure remains an important part of AI deployment. #TRONEcoStar @Justin Sun孙宇晨