Example of a reward webservice exposing the Ushiriki Policy Engine to the API. The general framework uses blueprints to structure the existence of optional environments.
This code has been tested with Python3, although if previous versions are required, the code content should suffice. It is written in a modular manner so environments can be added at any time. When they are added, the yaml file must be revised to point to the module path of the new objects.
First clone or download the soure from the repository:
git clone https://github.com/IBM/ushiriki-policy-engine-reward-service
Ensure that all the required modules are included into the base version of requirements.txt
(To get all the dependencies you might have to check in your subdirectories
cat */requirements >> requirements.txt)
pip install -r requirements.txt
Selecting the appropriate configuration file defining the environment you'd like to host..
PORT=8080 python webservice.py config.yaml
[Environment Reward Service]
@misc{pending,
Author = {Sekou L Remy and Oliver E Bent},
Title = {A Global Health Gym Environment for RL Applications},
Year = {2020},
Eprint = {arXiv:pending},
}