colour.utilities.as_ndarray#

colour.utilities.as_ndarray(a: Any) → np.ndarray[source]#

Convert the specified array \(a\) to a numpy.ndarray.

This function handles arrays from any backend (JAX, PyTorch, CuPy, etc.) by moving them to the host when direct conversion is not possible, e.g., for device-resident arrays.

Parameters:

a (Any) – Array, scalar, or Python sequence to convert.

Returns:

NumPy array.

Return type:

numpy.ndarray

Notes

  • Unlike as_array() / as_float_array() siblings, as_ndarray() does not honour _NDARRAY_COPY_ENABLED. It is a backend-host hand-off boundary helper, not a copy toggle; the returned array shares storage with the input wherever the backend allows.

Examples

>>> import numpy as np
>>> as_ndarray(np.array([1, 2, 3]))
array([1, 2, 3])

Round-trip a PyTorch tensor on a non-host device:

>>> import torch
>>> as_ndarray(torch.tensor([1, 2, 3], device="mps"))
array([1, 2, 3])
On this page

This Page