[Enhancement] Add the dtype limit of nms_npu to maintain consistency with the GPU (#2724)

* Increase the dtype limit to maintain consistency with the gpu.

* update error message
Co-authored-by: Zaida Zhou <58739961+zhouzaida@users.noreply.github.com>

---------

Co-authored-by: momo609 <963372609.qq.com>
Co-authored-by: Zaida Zhou <58739961+zhouzaida@users.noreply.github.com>

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

@@ -4,6 +4,8 @@ using namespace NPU_NAME_SPACE;
 using namespace std;
 
 Tensor nms_npu(Tensor boxes, Tensor scores, float iou_threshold, int offset) {
+  TORCH_CHECK((boxes.scalar_type == at::ScalarType::Float),
+              "The type of boxes tensor passed in nms_npu should be float");
   int64_t offset_64 = offset;
   at::Tensor iou_threshold_y = at_npu::native::OpPreparation::ApplyTensor(
                                    {}, boxes.options().dtype(at::kFloat), boxes)