Contract-first recipes for hypercube-cnn. Not installed with the wheel — clone the repo and install the package, then run from any directory.
cd HypercubeCNN
pip install .
python examples/python/synthetic_classification.py
python examples/python/synthetic_regression.py
python examples/python/arch_and_weights_io.py
python examples/python/spatial_embed_smoke.py| Script | Checks |
|---|---|
synthetic_classification.py |
Train a few epochs; accuracy improves |
synthetic_regression.py |
Train a few epochs; MSE drops |
arch_and_weights_io.py |
export_arch + HCNW round-trip; bit-identical logits |
spatial_embed_smoke.py |
Spatial pad_value contract, DualPlane, eval helpers |
Kept small and offline (no dataset download). Full API: docs/Python_SDK.md.