Fix compilation error (pytorch 1.11.0, cuda 11.6)

[sitic]

Compilation failed for src/spmm.cu with

/usr/include/c10/util/ArrayRef.h(142): error: expression must be a pointer to a complete object
 type
          detected during instantiation of "c10::ArrayRef<T>::iterator c10::ArrayRef<T>::end()
const [with T=at::Tensor]"
/usr/include/ATen/cuda/CUDAUtils.h(14): here

/usr/include/ATen/cuda/CUDAUtils.h(15): error: incomplete class type "at::Tensor" is not allowed

/usr/include/ATen/NamedTensorUtils.h(14): error: no instance of function template "std::any_of"
 matches the argument list
            argument types are: (const at::Tensor *, <error-type>, lambda [](const at::Tensor &
)->__nv_bool)

for pytorch 1.11.0 with cuda 11.6.

Fixes #453

[StefOe]

this patch worked for me (pytorch 1.12 cuda 11.3)

[chrischoy]

Thanks for the PR!

[123sleaf-123]

You are the real hero, sir. Thank you!!

But I'm confused that why I can't see your change in the master branch? (I add this line of code by myself)