* add unit tests for the two versions of Backend create_tensor

* disable new unit test on GPU until we have time to address it