Change to stable sort in nms implementations (#4767)

* change to stable sort in nms implementations

torchvision/csrc/ops/cpu/nms_kernel.cpp

@@ -27,7 +27,8 @@ at::Tensor nms_kernel_impl(
 
   at::Tensor areas_t = (x2_t - x1_t) * (y2_t - y1_t);
 
-  auto order_t = std::get<1>(scores.sort(0, /* descending=*/true));
+  auto order_t = std::get<1>(
+      scores.sort(/*stable=*/true, /*dim=*/0, /* descending=*/true));
 
   auto ndets = dets.size(0);
   at::Tensor suppressed_t = at::zeros({ndets}, dets.options().dtype(at::kByte));

torchvision/csrc/ops/cuda/nms_kernel.cu

@@ -109,7 +109,8 @@ at::Tensor nms_kernel(
     return at::empty({0}, dets.options().dtype(at::kLong));
   }
 
-  auto order_t = std::get<1>(scores.sort(0, /* descending=*/true));
+  auto order_t = std::get<1>(
+      scores.sort(/*stable=*/true, /*dim=*/0, /* descending=*/true));
   auto dets_sorted = dets.index_select(0, order_t).contiguous();
 
   int dets_num = dets.size(0);

torchvision/csrc/ops/quantized/cpu/qnms_kernel.cpp

@@ -27,7 +27,8 @@ at::Tensor qnms_kernel_impl(
   auto y1_t = dets.select(1, 1).contiguous();
   auto x2_t = dets.select(1, 2).contiguous();
   auto y2_t = dets.select(1, 3).contiguous();
-  auto order_t = std::get<1>(scores.sort(0, /* descending=*/true));
+  auto order_t = std::get<1>(
+      scores.sort(/*stable=*/true, /*dim=*/0, /* descending=*/true));
   at::Tensor suppressed_t = at::zeros({ndets}, dets.options().dtype(at::kByte));
   at::Tensor keep_t = at::zeros({ndets}, dets.options().dtype(at::kLong));
   at::Tensor areas_t = at::zeros({ndets}, dets.options().dtype(at::kFloat));