[Intel GPU] Use user-friendly err msg in mm (pytorch#151655)

Pull Request resolved: pytorch#151655
Approved by: https://github.com/EikanWang

Author: ZhiweiYan-96

aten/src/ATen/native/mkldnn/xpu/Blas.cpp

@@ -154,7 +154,7 @@ Tensor& mm_out(const Tensor& self, const Tensor& mat2, Tensor& result) {
       ")");
   TORCH_CHECK(
       self.dtype() == mat2.dtype(),
-      "expected self and mat2 to have the same dtype, but got: ",
+      "expected mat1 and mat2 to have the same dtype, but got: ",
       self.dtype(),
       " != ",
       mat2.dtype())