Skip to content

Installation

Requirements

PGTG currently requires Python 3.11, 3.12, or 3.13.

All dependencies are fetched and installed automatically.

Install from PyPI

PGTG is published on PyPI and can be installed using the package manager of your liking:

pip install pgtg

uv pip install pgtg

Install from source

Clone the repository and install it in editable mode:

git clone https://github.com/neuro-mechanistic-modeling/pgtg.git
cd pgtg
python -m venv .venv && source .venv/bin/activate
pip install -e .

git clone https://github.com/neuro-mechanistic-modeling/pgtg.git
cd pgtg
uv venv
uv pip install -e .

Optional extras

The optional dependency groups are exposed as extras:

Extra Installs Use for
docs MkDocs Material + mkdocstrings + ruff Building this documentation
dev pytest, pre-commit, ruff, marimo Running tests and the pre-commit hooks
examples marimo, Stable-Baselines3, torch, PyDSMC Running the example notebooks
all everything above A complete development environment

For development, we recommend installing all extras with:

pip install -e ".[all]"

The repository ships a mise configuration that pins versions and provides tasks for the common workflows, such as:

mise run install # create the venv and install the .[all] extra
mise run test    # run the test suite
mise run docs    # build the documentation (add --serve for live-reload)
mise run notebook <example-notebook>  # defaults to explore_pgtg.py

See Contributing for the full development workflow.

Verify the installation

import gymnasium as gym
import pgtg

gym.register_envs(pgtg)  # no-op; registering as an import side-effect

print(pgtg.__version__)
env = gym.make("pgtg-v5")
print(env.reset(seed=0)[0].keys())  # dict_keys(['position', 'velocity', 'map'])

Importing registers the environment

PGTG registers its environments as an import side-effect. You have to import pgtg (even if you never reference it directly). You can call gym.register_envs(pgtg) to prevent your formatter from removing the unused import.