OpenAI
1. Gym
Homepage
Github
Docs
Open AI. Greg Brockman, Vicki Cheung, Ludwig Pettersson, etc. 2016. URL: https://arxiv.org/abs/1606.01540
1.1 Install
1
2
3
4
| conda create -n gym python=3.5
source active gym
pip install gym
# pip install gym[all]
|
1.2 Environments
1
2
3
4
5
6
7
8
9
10
11
12
| import gym
env = gym.make('CartPole-v0')
for i_episode in range(20):
observation = env.reset()
for t in range(100):
env.render()
print(observation)
action = env.action_space.sample()
observation, reward, done, info = env.step(action)
if done:
print("Episode finished after {} timesteps".format(t+1))
break
|
1.3 Register
It’s very easy to add your own environments to the registry, and thus make them available for $gym.make()$: just $register()$ them at load time.
1.4 Demo
1
2
3
4
5
6
7
8
9
10
11
| import gym
env=gym.make('CartPole-v0')
env=env.unwrapped
print(env.action_space)
print(env.observation_space)
print(env.observation_space.high)
print(env.observation_space.low)
printt(env.action_space.n)
print(env.observation_space.shape[0])
|