Giters

Kaixhin / PlaNet

Deep Planning Network: Control from pixels by latent planning with learned dynamics

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Test

deeplearnerJHB opened this issue · comments

commented

Thank you for your answer. I had found the bugs in my code that leads bad results in many environment. I also found you use the same seed for all 10 test environments and keep exploration in test phase. I'm confused with such operation. May be such operation leads the quick convergence in “reacher-easy” and huge variance in “ball_in_cup-catch”?