Installation Guide¶

Welcome to Science‑Gym — a suite of Gym-compatible environments for data‑driven scientific discovery. This guide shows the quickest way to get the library running on your machine, plus extra tips for GUI physics, Box2D, or symbolic‑regression extras.

Note

Heads‑up – Science‑Gym is still young and pulls in quite a few heavy dependencies (PyTorch, Stable‑Baselines 3, Box2D, OpenCV, SymPy…). We are actively refactoring to make the core install lighter and to split GPU/GUI tooling into optional extras. Thanks for your patience — stay tuned for slimmer wheels in the next minor release!

Prerequisites¶

Requirement	Recommended
Python	3.9 – 3.12 (64‑bit)
`pip` / `wheel`	Latest (`python -m pip install -U pip wheel`)
SWIG (needed only for Box2D)	4.1 +

Installing SWIG¶

Some environments rely on Box2D, which in turn needs SWIG to build its C++ bindings if a pre‑built wheel is not available for your platform.

OS	Command
Ubuntu / Debian	`sudo apt-get install swig`
macOS (Homebrew)	`brew install swig`
Windows (Chocolatey)	`choco install swig` or download the binary installer from swig.org and add it to `PATH`

If you do not plan to use the Box2D‑based tasks you can skip SWIG altogether.

Basic install¶

pip install science-gym

This installs the core library plus all other requirements. Sorry that it takes a while!

deterministic physics environments that do not require Box2D
vectorised “sandbox” tasks for regression and optimisation
the shared simulation / problem interface

Optional extras¶

# Reinforcement‑learning agents + PyTorch
pip install science-gym[rl]

# GUI & 2‑D physics (Box2D, OpenCV, Pygame)
pip install science-gym[gui]

# Symbolic regression + SymPy
pip install science-gym[sym]

# Everything (WARNING: heavy!)
pip install science-gym[all]

Extras can be mixed, e.g. pip install science-gym[rl,gui]. Note that, as of version 0.1, all of these packages are anyways installed with the basic pip install.

Bleeding‑edge / development version¶

git clone https://github.com/Pibborn/science-gym.git
cd science-gym
pip install --editable .[all]   # or choose the extras you need

The --editable flag lets you hack on the source and import the package without reinstalling.

Verifying your install¶

Put the following in any python script:

env = Sim_Lagrange()
input_dim, output_dim = get_env_dims(env)
train_problem = Problem_Lagrange(env)
print(output_dim)

Expected output:

Observation shape: (3,)

Congratulations — Science‑Gym is ready!

Feel free to open a GitHub issue if you bump into something unexpected.