MAPPO Attention Project Repository

Welcome to the repository for our MAPPO (Multi-Agent Proximal Policy Optimization) project. This repository contains the necessary code and scripts to run and manage MAPPO with modular attention architectures experiments.

Package Versions

The specific versions of the packages and dependencies used in our MAPPO implementation are detailed in the run_mppo_attention.sh script. This script is tailored for execution in a Compute Canada environment, ensuring a consistent and reproducible setup.

Accessing Package Version Information

To view the package versions, please refer to the run_mppo_attention.sh file located in the root directory of this repository. The script includes commands to install or load the specific versions of each required package.

Running the Script

If you have access to a Compute Canada environment, you can directly execute the run_mppo_attention.sh script to set up the environment and run the MAPPO experiments. Make sure you have the necessary permissions and environment modules loaded.

Steps to Run:

1. Clone the repository to your Compute Canada workspace:

   git clone https://github.com/neuronphysics/MAPPO-ATTENTIOAN.git developer

2. Navigate to the repository directory:

   cd onpolicy

3. Make the script executable (if not already):

   chmod +x run_mppo_attention.sh

4. Execute the script:

   ./run_mppo_attention.sh