README.md

Multi-Agent Marginal Q-Learning from Demonstrations

Official code base for MAMQL: Multi-Agent Marginal Q-Learning from Demonstrations

MAMQL is a sample efficienct multi-agent inverse reinforcement learning algorithm that achieves state of the art average reward and reward recovery on several benchmarks.

Install Dependencies

1. Install Python 3.10 or a more recent version.

2. Clone the MAMQL repo and install the Python packages in requirements.txt. You can use the following bash command:

   $ pip install -r requirements.txt

3. Clone and install other environment packages to run intersection and overcooked.

   $ git clone https://github.com/HumanCompatibleAI/overcooked_ai.git
   $ pip install -e overcooked_ai/
   $ pip install highway-env

4. Install PyTorch. The version of PyTorch.

   $ pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

Training and Running

1. To train MAMQL on environment run the main file with args to select environment. An example train command for the intersection environment:

   $ python main.py --setup 4 --config_path intersection\configs\setup2.json --env_name intersection

2. To test your model and generate gifs of the results, update the corresponding train_env file with the path name for the model and run the following command (for the same intersection env):

   $ python main.py --setup 4 --config_path intersection\configs\setup2.json --env_name intersection --mode test

Gems Environment Results

Intersection Environment Results

Overcooked Environment Results