SpyRL is a reinforcement learning (RL) framework that provides commonly used library, so you will write less code. For example, it automatically creates a learning graph for every learning session.
Use the package manager pip to install SpyRL.
pip install spyrl$ cd to project directory
$ pip install -r requirements.txt
$ pip install -e .Download and install MPI from https://www.microsoft.com/en-us/download/details.aspx?id=57467 (install both the .exe and .mpi files)import gym
import sys
from spyrl.activity.learning import Learning
from spyrl.activity.activity_config import ActivityConfig
from spyrl.listener.impl.basic_functions import BasicFunctions
from spyrl.agent_builder.impl.actor_critic_traces_agent_builder import ActorCriticTracesAgentBuilder
from example.cartpole.helper.cartpole_discretiser import CartpoleDiscretiser
from example.cartpole.helper.env_wrapper import GymEnvWrapper
if __name__ == '__main__':
env = gym.make('CartPole-v0')
num_actions = env.action_space.n
config = ActivityConfig(num_episodes=1000, out_path='result/cartpole/test1/')
agent_builder = ActorCriticTracesAgentBuilder(num_actions, discretiser=CartpoleDiscretiser())
learning = Learning(listener=BasicFunctions(render=False))
learning.learn(env, agent_builder, config)Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Please make sure to update tests as appropriate.
To cite this repository in publications:
@misc{spyrl,
author = {Kurniawan, Budi and Vamplew, Peter and Papasimeon, Michael and Dazeley, Richard and Foale, Cameron},
title = {SpyRL Framework},
year = {2021},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/budi-kurniawan/spyrl}},
}
...