colour.utilities.as_float_array

colour.utilities.as_float_array(a: ArrayLike, dtype: Type[DTypeFloat] | None = None) → NDArrayFloat

Convert the specified variable \(a\) to a floating-point array using the specified numpy.dtype.

This is a namespace-aware boundary helper. When Array API dispatch is enabled and \(a\) is a non-NumPy array (e.g. JAX, PyTorch), the result is returned in \(a\)’s native namespace, on its device, and cast to dtype. Otherwise the result is a numpy.ndarray. This is the convention used at function boundaries: a = as_float_array(a) followed by xp = array_namespace(a) recovers the caller’s backend.

Parameters:

• a (ArrayLike) – Variable \(a\) to convert.

• dtype (Type[DTypeFloat] | None) – Floating-point numpy.dtype to use for conversion, default to the numpy.dtype defined by the colour.constant.DTYPE_FLOAT_DEFAULT attribute.

Returns:

Variable \(a\) converted to a floating-point array in the input’s namespace.

Return type:

numpy.ndarray or backend tensor

Examples

>>> as_float_array([1, 2, 3])
array([1., 2., 3.])