How to enable or disable the NumPy backend

MARPLE auto-detects array backends at startup in this order: CuPy, NumPy, ulab. If none are found, it falls back to pure Python.

Use NumPy (recommended for desktop)

pip install numpy
marple

NumPy provides significant speedups for element-wise arithmetic, reduce, and inner product on large arrays.

Force pure Python

MARPLE_BACKEND=none marple

Set the MARPLE_BACKEND environment variable to none to disable all backends. This is useful for testing or when you want predictable Python-only behaviour.

Auto-detection (default)

When MARPLE_BACKEND is unset or set to auto, MARPLE tries CuPy first (for GPU acceleration), then NumPy, then ulab (for CircuitPython/MicroPython), and finally falls back to pure Python lists.

See also: Check which backend is active, Writing fast code