[Fix] Keep the shape of iou op's parameter 2 smaller than parameter 1 (#2821)

mmcv/ops/csrc/pytorch/npu/bbox_overlaps_npu.cpp

@@ -12,11 +12,17 @@ void bbox_overlaps_npu(const Tensor bboxes1, const Tensor bboxes2, Tensor ious,
   if (mode == 1) {
     modeStr = "iof";
   }
+  bool swap_flag = false;
   at::Tensor bboxesFP32 = bboxes2;
   at::Tensor gtboxesFP32 = bboxes1;
+  if (bboxes2.size(0) < bboxes1.size(0)) {
+    swap_flag = true;
+    bboxesFP32 = bboxes1;
+    gtboxesFP32 = bboxes2;
+  }
   if (bboxes2.scalar_type() != at::ScalarType::Float) {
-    bboxesFP32 = NPUNativeFunctions::npu_dtype_cast(bboxes2, at::kFloat);
-    gtboxesFP32 = NPUNativeFunctions::npu_dtype_cast(bboxes1, at::kFloat);
+    bboxesFP32 = NPUNativeFunctions::npu_dtype_cast(bboxesFP32, at::kFloat);
+    gtboxesFP32 = NPUNativeFunctions::npu_dtype_cast(gtboxesFP32, at::kFloat);
   }
   c10::SmallVector<int64_t, SIZE> iousSize = {gtboxesFP32.size(0),
                                               bboxesFP32.size(0)};
@@ -38,6 +44,7 @@ void bbox_overlaps_npu(const Tensor bboxes1, const Tensor bboxes2, Tensor ious,
   if (bboxes2.scalar_type() != at::ScalarType::Float) {
     iousFP32 = NPUNativeFunctions::npu_dtype_cast(iousFP32, at::kHalf);
   }
+  iousFP32 = swap_flag ? iousFP32.transpose(0, 1) : iousFP32;
   ious.copy_(iousFP32);
 }