/gym_x

Gym environments for capture properties of hidden states(hx) of recurrent networks.

Primary LanguagePython

gym_x

Environments designed to test various aspects of recurrent network policies. Their minimal state representation could be found here.

Installation

cd gym_x
pip install -e .

Usage:

>>> import gym
>>> import gym_x
>>> env = gym.make('GoldRushRead-v0')
>>> env.action_space
Discrete(4)
>>> env.reset()
array([1., 1.])
>>> env.step(env.action_space.sample())
(array([0., 0.]), 0, False, {'desired_action': 3})

Environments