Open Sohojoe opened 5 years ago
Thanks for pointing these issues out, @Sohojoe.
@ervteng Could you take a look and make the appropriate changes?
@Sohojoe thanks for pointing out these issues!
For the config, I've put it to more than 10M intentionally, since depending on environment (e.g. for a fixed seed) the agent may continue to learn well past 10M. I erred on the side of too many steps, as you can always kill a run but it would be very disappointing to wake up in the morning and find that your run has stopped prematurely! I've added a blurb in the documentation describing how to change this.
@ervteng that sounds good - I think the main thing is to have it clear in the guide
From GCP tutorial
[bug] The tutorial references an old release
should be
[bug] Install the Obstacle Tower Environment is missing
cd obstacle-tower-env
should be
[recomendation/discussion] I believe the config will run for 50m steps by default - maybe the tutorial should override this with 10m steps (as this seems to be what was recommended)