Installation¶

KoopmanGraph requires Python 3.10 or newer, PyTorch, and PyTorch Geometric (PyG). Install PyTorch and PyG before installing KoopmanGraph so pip can resolve compatible wheels for your platform.

Prerequisites¶

Python¶

Use Python 3.10, 3.11, or 3.12. Check your version:

python --version

PyTorch¶

Install the PyTorch build that matches your system (CPU or CUDA). Follow the selector at PyTorch Get Started. Example for CPU-only on Linux or macOS:

pip install torch

PyTorch Geometric¶

PyG depends on the PyTorch version already installed. Use the official PyG installation guide to pick the correct torch-geometric (and optional extension) wheels.

Minimal install after PyTorch is in place:

pip install torch-geometric

Install KoopmanGraph¶

From source (recommended)¶

Clone the repository and install in editable mode:

git clone https://github.com/tjkessler/KoopmanGraph.git
cd KoopmanGraph
pip install -e .

For development (tests, linting, pre-commit):

pip install -e ".[dev]"

To build documentation locally:

pip install -e ".[docs]"
cd docs && make html

PyPI¶

After PyTorch and PyG are installed, install KoopmanGraph from PyPI:

pip install koopman-graph

Pin a specific release when reproducing results:

pip install koopman-graph==0.2.0

Releases are published automatically when a maintainer creates a GitHub Release (see CONTRIBUTING.md in the repository). Until the first release is published, use the editable install from source above.

Verify¶

Confirm the package imports:

python -c "import koopman_graph; print(koopman_graph.__version__)"

For a full development check after pip install -e ".[dev]":

pytest tests/ -v

Next steps¶

See Quickstart for a minimal train-and-predict workflow.