Installation

Requirements

ARCADE targets Python 3.10 or newer. The core install pulls numerical and scientific dependencies used by nearly every path: NumPy, SciPy, h5py, scikit-learn, Matplotlib, Seaborn, PyYAML, Pydantic, and tqdm. Neural methods need PyTorch. Path signature features need iisignature. FPGA-oriented export helpers need hls4ml. Quantum error correction curves need Stim, Sinter, and PyMatching. Documentation builds need Sphinx and the Read the Docs theme. Development installs add pytest and ruff.

You do not need every optional stack on day one. Install the core package first, confirm the command line works, then add extras as you enable the corresponding YAML sections.

Create an environment and install

From the repository root:

cd ARCADE
python -m venv .venv
source .venv/bin/activate
pip install -e ".[torch,signature]"
bash scripts/link_paper_data.sh

The editable install links the arcade package into the environment so local edits apply without a reinstall. The torch extra enables neural classifiers such as FNN, HERQULES, leakage, and multilevel networks. The signature extra enables the path signature classifier. The link_paper_data.sh script creates the data/ symlinks that bundled paper configs expect for the public Readout 2019 HDF5 files.

Optional extras

Extra	Install command	What it unlocks
Core only	`pip install -e .`	Threshold and scikit-learn baselines
`torch`	`pip install -e ".[torch]"`	Neural classifiers
`signature`	`pip install -e ".[signature]"`	Path signature method
`hardware`	`pip install -e ".[hardware]"`	HLS4ML-oriented helpers
`qec`	`pip install -e ".[qec]"`	Stim-based LER and FTQC curves
`dev`	`pip install -e ".[dev]"`	pytest and ruff
`docs`	`pip install -e ".[docs]"`	Sphinx build via `make -C doc html`
`export`	`pip install -e ".[export]"`	PDF figure export with PyMuPDF
`all`	`pip install -e ".[all]"`	Full dependency set

For a documentation-only contributor who will not train networks, pip install -e ".[docs]" is enough to build this manual. For a zoo run that matches the public multi-method benchmark, install at least torch and signature. For QEC pages and configs, add qec.

Command line entry point

The project exposes a console script named arcade. After a successful editable install you should be able to run:

arcade list
arcade describe threshold

arcade list prints registered classifier names. arcade describe prints a short description string from the registry. The full pipeline entry point is arcade run path/to/config.yaml, which loads configuration, runs stages in order, and writes outputs under the directory named in the visualization section.

Building this documentation

pip install -e ".[docs]"
make -C doc html
python -m http.server -d doc/_build/html 8000

Open http://127.0.0.1:8000/ in a browser. The HTML theme is the Read the Docs theme chosen for a fixed, reference-manual look with low visual noise.

Next

Continue with Pipeline stages for a stage-by-stage map of what happens after arcade run. Then use Verification to smoke-test the environment.