[cherry-pick] add multiclass_nms2 kernel, test=develop (#2917) (#2944)

add multiclass_nms2 kernel, test=develop

lite/kernels/host/multiclass_nms_compute.cc

@@ -429,6 +429,17 @@ REGISTER_LITE_KERNEL(multiclass_nms,
     .BindInput("BBoxes", {LiteType::GetTensorTy(TARGET(kHost))})
     .BindInput("Scores", {LiteType::GetTensorTy(TARGET(kHost))})
     .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kHost))})
+    .Finalize();
+
+REGISTER_LITE_KERNEL(multiclass_nms2,
+                     kHost,
+                     kFloat,
+                     kNCHW,
+                     paddle::lite::kernels::host::MulticlassNmsCompute,
+                     def)
+    .BindInput("BBoxes", {LiteType::GetTensorTy(TARGET(kHost))})
+    .BindInput("Scores", {LiteType::GetTensorTy(TARGET(kHost))})
+    .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kHost))})
     .BindOutput("Index",
                 {LiteType::GetTensorTy(TARGET(kHost), PRECISION(kInt32))})
     .Finalize();

lite/operators/multiclass_nms_op.cc

@@ -84,3 +84,4 @@ bool MulticlassNmsOpLite::AttachImpl(const cpp::OpDesc& opdesc,
 }  // namespace paddle
 
 REGISTER_LITE_OP(multiclass_nms, paddle::lite::operators::MulticlassNmsOpLite);
+REGISTER_LITE_OP(multiclass_nms2, paddle::lite::operators::MulticlassNmsOpLite);