Entity Neural Network Zoo

The enn-zoo package collects entity-gym bindings for a number of reinforcement learning environments:

• Procgen
• Griddly
• MicroRTS
• ViZDoom
• CodeCraft

Setup

git clone https://github.com/entity-neural-network/enn-zoo.git
cd enn-zoo
poetry install
poetry run pip install setuptools==59.5.0
poetry run pip install torch==1.10.2+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
poetry run pip install torch-scatter -f https://data.pyg.org/whl/torch-1.10.0+cu113.html

Some of the environments have additional dependencies:

sudo apt install python3-dev make build-essential libssl-dev zlib1g-dev \
libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm \
libncursesw5-dev xz-utils tk-dev libxml2-dev libxmlsec1-dev libffi-dev liblzma-dev

Usage

poetry run python enn_zoo/train.py

Gym-µRTS

Gym-µRTS is a Reinforcement Learning environment for the popular Real-time Strategy game simulator μRTS. To get started, run the following command:

xvfb-run -a poetry run python enn_zoo/train.py \
    env.id=GymMicrorts \
    rollout.num_envs=24 \
    total_timesteps=1000000 \
    rollout.steps=256 \
    track=true \
    eval.capture_videos=True \
    eval.interval=300000 \
    eval.steps=2000 \
    eval.num_envs=1

Here is a tracked Gym-µRTS experiment, which has a trained agent that behaves as follows:

ezgif-3-d39df99e1a.mp4