Hello I have a question about using MKL and CBLAS.
In generic/THBlas.c, lines 320 and 322 call the BLAS functions dgemm_ and sgemm_.
I would like to substitute them with their CBLAS counterparts referenced here:
https://software.intel.com/en-us/mkl-developer-reference-c-cblas-gemm
I tried to include “mkl.h” in THBlas.c but the PyTorch setup does not recognize the header file. Without the header file the setup throws errors for CBLAS_LAYOUT and CBLAS_TRANSPOSE types.
The end goal is to use Sparse BLAS routines referenced here:
https://software.intel.com/en-us/mkl-developer-reference-c-mkl-csrmm
My PyTorch was installed with Anaconda which includes the MKL library.
What steps do I need to make the switch?