Implementation problem about Learning to Explore via Meta-Policy Gradient

[Shenqi-Li]

I implemented this paper, but it can not achieve paper's performance. In the teacher-student task, do you have some drift problem? Could you share your code?