Python Array API

Cubed implements version 2022.12 of the Python Array API standard in cubed.array_api, with a few exceptions listed on the coverage status page. The Fourier transform functions are not supported.

Support for version 2023.12 is tracked in Cubed issue #438.

Differences between Cubed and the standard

The following Creation Functions accept extra chunks and spec keyword arguments:

cubed.array_api.arange(start, /, stop=None, step=1, *, dtype=None, device=None, chunks='auto', spec=None)

cubed.array_api.asarray(obj, /, *, dtype=None, device=None, copy=None, chunks='auto', spec=None)

cubed.array_api.empty(shape, *, dtype=None, device=None, chunks='auto', spec=None)

cubed.array_api.empty_like(x, /, *, dtype=None, device=None, chunks=None, spec=None)

cubed.array_api.eye(n_rows, n_cols=None, /, *, k=0, dtype=None, device=None, chunks='auto', spec=None)

cubed.array_api.full(shape, fill_value, *, dtype=None, device=None, chunks='auto', spec=None)

cubed.array_api.full_like(x, /, fill_value, *, dtype=None, device=None, chunks=None, spec=None)

cubed.array_api.linspace(start, stop, /, num, *, dtype=None, device=None, endpoint=True, chunks='auto', spec=None)

cubed.array_api.ones(shape, *, dtype=None, device=None, chunks='auto', spec=None)

cubed.array_api.ones_like(x, /, *, dtype=None, device=None, chunks=None, spec=None)

cubed.array_api.zeros(shape, *, dtype=None, device=None, chunks='auto', spec=None)

cubed.array_api.zeros_like(x, /, *, dtype=None, device=None, chunks=None, spec=None)

The following Manipulation Functions accept extra chunks keyword arguments:

cubed.array_api.broadcast_to(x, /, shape, *, chunks=None)

cubed.array_api.concat(arrays, /, *, axis=0, chunks=None)