Readme.md

Integration with stable-baselines algorithms example

This example shows how to use stable-baselines, a compilation of state of the art RL algorithms made for OpenAI Gym, in ARLIE.

Additional requirements

See detailed information on installing stable-baselines here.

Stabe-Baselines supports Tensorflow versions from 1.8.0 to 1.14.0. Support for Tensorflow 2 API is planned.

pip install tensorflow==1.14

Or with GPU support:

pip install tensorflow-gpu==1.14

Obviously, Stable baselines is required to run this example

pip install stable-baselines

You can include an optional dependency on MPI if you use it, enabling algorithms DDPG, GAIL, PPO1 and TRPO.

pip install stable-baselines[mpi]

Optionally, the example allows to run the same model using the LunarLander-v2 environment from gym. Additional requirements need to be installed to run the gym version:

In windows, you will first need to download and install Build Tools for Visual Studio. Then restart and:

conda install swig
pip install gym[box2d]