How to support array devices? #65
Description
Is your feature request related to a problem? Please describe.
We have many functions which use array constructions such as xp.ones(shape), which is problematic if an input is e.g. torch.ones(shape, device="cuda") and torch default device is cpu. Should we be reading the device as well as the array_backend from the inputs and if so, how should it be handled when inputs are not on the same device?
Describe the solution you'd like
We should be able to handle torch inputs regardless of device
Describe alternatives you've considered
xp.ones_like(arr) could implicitly handle this also if we have a reference array we can use
Additional context
No response
Organisation
No response