Fixes deform_conv issue with large input/output (#4351)

* WIP on fixing index overflow issue

* Fixed backward pass for large num_kernels

* Fixed clang formatting

* Fixed GET_BLOCKS int/int64_t types issue
Co-authored-by: vfdev-5 <vfdev-5@gmail.com>
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

torchvision/csrc/ops/cuda/cuda_helpers.h

@@ -3,10 +3,12 @@
 namespace vision {
 namespace ops {
 
-#define CUDA_1D_KERNEL_LOOP(i, n)                                \
-  for (int i = (blockIdx.x * blockDim.x) + threadIdx.x; i < (n); \
+#define CUDA_1D_KERNEL_LOOP_T(i, n, index_t)                         \
+  for (index_t i = (blockIdx.x * blockDim.x) + threadIdx.x; i < (n); \
        i += (blockDim.x * gridDim.x))
 
+#define CUDA_1D_KERNEL_LOOP(i, n) CUDA_1D_KERNEL_LOOP_T(i, n, int)
+
 template <typename integer>
 constexpr __host__ __device__ inline integer ceil_div(integer n, integer m) {
   return (n + m - 1) / m;