Fixed test_numba_integration (#25017)
Summary:
The semantic of the _auto-convert GPU arrays that support the __cuda_array_interface__ protocol_ has changed a bit.
It used to throw an exception when using `touch.as_tensor(...,device=D)` where `D` is a CUDA device not used in `__cuda_array_interface__`. Now, this is supported and will result in an implicit copy.
I do not what have changes but `from_blob()` now supports that the input and the output device differ.
I have updated the tests to reflect this, which fixes https://github.com/pytorch/pytorch/issues/24968
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25017
Differential Revision: D16986240
Pulled By: soumith
fbshipit-source-id: e6f7e2472365f924ca155ce006c8a9213f0743a7
