Testing

Hardware-aware test launcher

Run the same launcher used by local commit hooks and GitHub Actions:

python scripts/run_tests.py -q

When torch.cuda.is_available() returns true, the launcher runs the full suite. Otherwise it adds -m "not cuda" and excludes tests that require a CUDA CT projector.

Explicit selections

Run only CPU-safe tests:

pytest -q -m "not cuda"

Run only CUDA-marked tests:

pytest -q -m cuda

Run a focused module while developing:

pytest -q tests/models/test_padis_reconstructor.py

Continuous integration

The Tests GitHub Actions workflow invokes the hardware-aware launcher. A standard ubuntu-latest worker runs the CPU suite; routing the job to a properly configured GPU runner automatically includes CUDA tests. The status is reported in the README but is not configured as a mandatory merge check.