Run OpenAI Gym environments on an external process or remote machine using gRPC.
Installation
Install Gym (not required if using Docker) and run:
pip install gymx
It is recommended to use a virtual environment.
Usage
Server
To start the server run:
python -m gymx
To use a different port run:
python -m gymx --port=54321
You can also run the server using Docker:
docker run -p 54321:54321 album/gymx
Client
Inside your application use:
from gymx import Env
env = Env('CartPole-v0')
To specify the server address use:
env = Env('CartPole-v0', address='localhost:54321')
API
-
env.reset()
: Reset the environment's state. Returnsobservation
. -
env.step(action)
: Step the environment by one timestep. Returnsobservation
,reward
,done
,next_episode
. Unlike the original gym API, it automatically resets the environment when done and returns next episode's observation instead ofinfo
.