前往新版Gitcode,体验更适合开发者的 AI 搜索 >>
Created by: jacquesqiao
First, instead of declaring a Trainer and call trainer.train(), is there any way to call the train_op only once (aka stepping)