Test Your Installation¶

Quick test¶

Run the built-in test on a GPU-capable machine:

recovar run_test_dataset

This generates a small synthetic dataset and runs the full pipeline. It should complete in a few minutes and confirms that JAX, CUDA, and all dependencies are working.

Development test suite¶

If you installed from source, run the test suite:

# Fast unit tests (no GPU required, ~30 seconds)
./scripts/run_pytests.sh fast

# Pipeline smoke test (CPU only, ~2 min)
./scripts/run_pytests.sh smoke

# GPU tests (requires GPU)
./scripts/run_pytests.sh gpu

# Full suite (unit + integration + GPU + slow, ~1-2h)
./scripts/run_pytests.sh full

# Quick quality check (small synthetic data, ~30 min, GPU)
./scripts/run_pytests.sh tiny-metrics

# Full quality regression (synthetic data, 6-12h, GPU)
./scripts/run_pytests.sh long-test

Or via pixi:

pixi run test-fast          # unit tests only
pixi run test-full          # full suite
pixi run test-parallel      # parallel Slurm submission
pixi run test-parallel-long # parallel including long regression

All test data is generated synthetically — no external downloads required.

See tests/README.md for test layout, markers, baselines, and guidelines.

Common issues¶

JAX not finding GPU¶

No GPU found. Set --accept-cpu if you really want to run on CPU

Check that JAX can see your GPU:

import jax
print(jax.devices())  # Should show CudaDevice

If it shows only CPU devices, reinstall JAX with CUDA support:

pip install "jax[cuda12]"==0.9.0.1

CUDA driver version mismatch¶

If you get CUDA errors, ensure your driver version is compatible with the CUDA toolkit version used by JAX. Check with:

nvidia-smi  # Shows driver version
python -c "import jax; print(jax.lib.xla_bridge.get_backend().platform_version)"