Thanks for your great work.
I had gym 0.7.3 installed before and I can use the env attributes like "x_threashold" in Cartpole.
But now, in gym 0.7.4 and higher, I cannot use those anymore.
Also, in MountainCar, LunnarLander and possibly more, the episode seems to be limited to some time range. Hope that be fixed soon.
If you're trying to munge attributes of the environment (which renders your learning results not comparable with anyone else's) you can get at the raw environment with env.unwrapped, as in env.unwrapped.x_threshold = -1
The episode is indeed limited to a time range, by design. An environment with unlimited time isn't as challenging to solve.
Thanks
On 20 March 2017 at 06:18, Trevor Blackwell notifications@github.com
wrote:
Closed #537 https://github.com/openai/gym/issues/537.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/openai/gym/issues/537#event-1005955234, or mute the
thread
https://github.com/notifications/unsubscribe-auth/ASgmNHX1WzJqDm7dvKW9YX6aSjjoL918ks5rnX9qgaJpZM4Mhy3B
.
Most helpful comment
If you're trying to munge attributes of the environment (which renders your learning results not comparable with anyone else's) you can get at the raw environment with
env.unwrapped, as inenv.unwrapped.x_threshold = -1The episode is indeed limited to a time range, by design. An environment with unlimited time isn't as challenging to solve.