search

FluxML / Gym.jl

Gym environments in Julia
MIT License
54 stars 19 forks source link

Register? #43

Open DhairyaLGandhi opened 4 years ago

darsnack commented 4 years ago

Would it make sense to use https://github.com/JuliaPOMDP/RLInterface.jl before registering this? That would make the environments interface with the rest of JuliaPOMDP which has a lot of policies and methods already implemented. I could see if there is interest in the community coordination efforts to work on this.

I can work on this package starting June-ish.

avik-pal commented 3 years ago

Maybe we should be using https://github.com/JuliaReinforcementLearning/CommonRLInterface.jl which would allow using the JuliaReinforcementLearning policies in addition to POMDP ones.

darsnack commented 3 years ago

Yup I was thinking the same thing. Would be happy to give it a shot. Maybe this weekend?

avik-pal commented 3 years ago

From the discussion there it seems that most of the functionality of this package is already provided in RLEnv (and they work pretty well from my experience), only it doesn't have headless rendering for image-based observation spaces. Makes me wonder why do we need this package specifically, and not just contribute the rendering feature there.

darsnack commented 3 years ago

Again was thinking along the same lines when I was commenting on that issue. RLEnv implements an interface that supersedes CommonRLInterface. I was thinking of doing the later here as a "test." Then porting the headless stuff over the RLEnv for the future.