This package contains basic tools to test consensus-based on- and off-policy multi-agent actor-critic algorithms on a couple simple environments. It was used to perform the empirical evaluation in https://arxiv.org/pdf/1903.06372.pdf.
Once you've cloned the repository, simply cd into it and do pip install -e ..
The code in this repository is a work-in-progress and is not yet well-documented. It will likely be massively overhauled in the future with complete disregard for backward compatibility.