Make compatible with pytorch 1.11 and newer; bugfix (#418)

* Bugfix: missing include common.h Signed-off-by: Krishna Murthy <krrish94@gmail.com> * Make compatible with pytorch 1.11 and newer -- THCudaCheck() deprecated Signed-off-by: Krishna Murthy <krrish94@gmail.com> --------- Signed-off-by: Krishna Murthy <krrish94@gmail.com>
zhanghang1989 · Feb 8, 2023 · c959dab · c959dab
1 parent 331ecdd
commit c959dab
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 8 deletions.
diff --git a/encoding/lib/gpu/lib_ssd.cu b/encoding/lib/gpu/lib_ssd.cu
@@ -19,8 +19,6 @@
 
 #include <ATen/ATen.h>
 #include <ATen/cuda/CUDAContext.h>
-#include <THC/THCNumerics.cuh>
-#include <THC/THC.h>
 
 #include <cuda.h>
 
@@ -378,22 +376,22 @@ std::vector<at::Tensor> box_encoder(const int N_img,
   printf("allocating %lu bytes for output labels\n", N_img*M*sizeof(long));
 #endif
   at::Tensor labels_out = at::empty({N_img * M}, labels_input.options());
-  THCudaCheck(cudaGetLastError());
+  C10_CUDA_CHECK(cudaGetLastError());
 
   // copy default boxes to outputs
 #ifdef DEBUG
   printf("allocating %lu bytes for output bboxes\n", N_img*M*4*sizeof(float));
 #endif
   at::Tensor bbox_out = dbox.repeat({N_img, 1});
-  THCudaCheck(cudaGetLastError());
+  C10_CUDA_CHECK(cudaGetLastError());
 
   // need to allocate some workspace
 #ifdef DEBUG
   printf("allocating %lu bytes for workspace\n", 8*M*N_img);
 #endif
   // at::Tensor workspace = at::CUDA(at::kByte).zeros({8 * M * N_img});
   at::Tensor workspace = at::zeros({8 * M * N_img}, at::CUDA(at::kByte));
-  THCudaCheck(cudaGetLastError());
+  C10_CUDA_CHECK(cudaGetLastError());
 
   // Encode the inputs
   const int THREADS_PER_BLOCK = 256;
@@ -408,7 +406,7 @@ std::vector<at::Tensor> box_encoder(const int N_img,
                       (float4*)bbox_out.data_ptr<float>(),
                       labels_out.data_ptr<long>());
 
-  THCudaCheck(cudaGetLastError());
+  C10_CUDA_CHECK(cudaGetLastError());
   return {bbox_out, labels_out};
 }
 
@@ -435,7 +433,7 @@ at::Tensor calc_ious(const int N_img,
                         (float4*)boxes2.data_ptr<float>(),
                         ious.data_ptr<float>());
 
-  THCudaCheck(cudaGetLastError());
+  C10_CUDA_CHECK(cudaGetLastError());
   return ious;
 }
 
@@ -569,7 +567,7 @@ std::vector<at::Tensor> random_horiz_flip(
           flip.data_ptr<float>(),
           tmp_img.data_ptr<scalar_t>(),
           nhwc);
-        THCudaCheck(cudaGetLastError());
+        C10_CUDA_CHECK(cudaGetLastError());
       });
 
   // copy tmp_img -> img

diff --git a/encoding/lib/gpu/rectify_cuda.cu b/encoding/lib/gpu/rectify_cuda.cu
@@ -10,6 +10,8 @@
 #include <ATen/cuda/detail/IndexUtils.cuh>
 #include <ATen/cuda/detail/KernelUtils.h>
 
+#include "common.h"
+
 
 template <typename dest_t, typename src_t>
 static inline dest_t safe_downcast(src_t v)