This repo implements:
- A multi-agent gridworld environment in
gym_multigrid/ - Multi-agent PPO training code in
multiagent_tfagents/ - Adversarial environment generation code with PAIRED in
adversarial_env/
The latter directory implements the code supporting the publication:
Dennis, M.*, Jaques, N.*, Vinitsky, E., Bayen, A., Russell, S., Critch, A., & Levine, S., Emergent Complexity and Zero-Shot Transfer via Unsupervised Environment Design, Neural Information Processing Systems (NeurIPS), Virtual (2020).
Please see the README file in each subdirectory for more details.
This is not an official Google product.
-
requires linux
-
create Python 3.10 environment (e.g. with conda $ conda create -n adv python=3.10 =)
-
activate the environment
-
$ pip3 install --upgrade pip
-
$ pip3 install tensorflow
-
$ pip3 install -r requirements.txt
-
Run test: $ python3 -m adversarial_env.test_adversarial_env