43 lines
788 B
Nix
43 lines
788 B
Nix
{ lib
|
|
, buildPythonPackage
|
|
, fetchFromGitHub
|
|
, numpy
|
|
, requests
|
|
, pyglet
|
|
, scipy
|
|
, pillow
|
|
, cloudpickle
|
|
}:
|
|
|
|
buildPythonPackage rec {
|
|
pname = "gym";
|
|
version = "0.18.1";
|
|
|
|
src = fetchFromGitHub {
|
|
owner = "openai";
|
|
repo = pname;
|
|
rev = version;
|
|
sha256 = "0mv4af2y9d1y97bsda94f21nis2jm1zkzv7c806vmvzh5s4r8nfn";
|
|
};
|
|
|
|
propagatedBuildInputs = [
|
|
cloudpickle
|
|
numpy
|
|
pillow
|
|
pyglet
|
|
requests
|
|
scipy
|
|
];
|
|
|
|
# The test needs MuJoCo that is not free library.
|
|
doCheck = false;
|
|
|
|
pythonImportsCheck = [ "gym" ];
|
|
|
|
meta = with lib; {
|
|
description = "A toolkit for developing and comparing your reinforcement learning agents";
|
|
homepage = "https://gym.openai.com/";
|
|
license = licenses.mit;
|
|
maintainers = with maintainers; [ hyphon81 ];
|
|
};
|
|
}
|