Configuration
MENOS has no configuration file and reads no environment variables. Behaviour is controlled by per-call options and by build-time CMake settings.
Per-call options
Qiskit
| Option | Where | Default |
|---|---|---|
shots | MENOSBackend(...), run(...), set_options(...) | 1024 |
seed_simulator | same | None (fresh random) |
want_statevector | same | False |
default_precision | MENOSEstimator(...) | 0.0 (recorded only) |
See Qiskit Backend and Estimator.
Native
Field of RunOptions | Default |
|---|---|
shots | 0 |
want_statevector | False |
seed | 0 |
memory_budget_bytes | None (50% of physical memory) |
expectation | None |
time_budget_seconds | None (unmonitored) |
See Python API.
Build-time settings
| CMake setting | Default | Purpose |
|---|---|---|
MENOS_BUILD_PYTHON_BINDINGS | OFF (the Python package turns it ON) | Build the menos._core extension |
MENOS_ENABLE_CUDA | OFF | Build the CUDA dense backend |
MENOS_BUILD_BENCHMARKS | OFF | Build the Google Benchmark programs |
CMAKE_CUDA_COMPILER | unset | Path to nvcc. Set it explicitly when more than one toolkit is installed |
CMAKE_CUDA_ARCHITECTURES | compiler default | Set it explicitly to your GPU's compute capability, for example 86 |
CUDAToolkit_ROOT | unset | Root of the toolkit to link against |
CMake presets: dev, release, cuda-dev, cuda-release, cuda12-dev, cuda12-release. The cuda-* presets pin /usr/local/cuda, the cuda12-* presets pin Ubuntu's CUDA 12 toolkit (/usr/bin/nvcc) with GCC 13, and all of them pin compute capability 8.6. These values are which are specific to the development machine; override them on the command line for another machine. See CUDA Installation.
Passing settings to a Python install
CMAKE_ARGS="-DMENOS_ENABLE_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=86" \
uv pip install --python .venv/bin/python ".[qiskit]"Budgets
The defaults and semantics of the memory and time budgets are in Memory Constraints and Time Budgets.
Calibration
Calibration runs automatically on first use and cannot be configured. See Cost Model.