Skip to content

Testing

A small, fast test suite covers the core Hamiltonian → qubit → VQE pipeline:

pip install pytest
pytest atomqc/tests/

The tests in atomqc/tests/test_qiskit.py build a small synthetic Hermitian \(H(k)\), decompose it into Pauli strings, and check the solver against exact NumPy diagonalization. They run offline in a few seconds (no JARVIS downloads), which is also what CI runs on every push.

What the tests check

  • test_hermitian_solver_basics — HermitianSolver pads \(H(k)\) to a power of two and reports the right qubit count.
  • test_pauli_decomposition_roundtrip — decompose_Hamiltonian reproduces the original matrix from its Pauli-string expansion.
  • test_vqe_matches_numpy_ground_state — VQE on the statevector simulator respects the variational principle (\(E_\text{VQE} \ge E_\text{ground}\)) and lands near the exact ground state. The closeness bound is intentionally loose because the classical optimizer may settle in a shallow local minimum, which is not reproducible across machines.

Continuous integration

GitHub Actions builds the conda environment from environment.yml, installs the package, and runs the suite under coverage on every push and pull request. The workflow lives at .github/workflows/main.yml.