What is the difference between these two methods on tensors?
There should be no difference as ScalarType is the C++ side for dtype.
The reason I ask is because I notice that they have different return types:
I see they are convertible to each other, should we just always prefer scalar_type then?