diff --git a/paddle/operators/concat_op.h b/paddle/operators/concat_op.h
index bff453971a00a75c7c7a495021cb96bcbfcdaa1c..c113f19fb5cf806709bff845ee0f1078b34014bb 100644
--- a/paddle/operators/concat_op.h
+++ b/paddle/operators/concat_op.h
@@ -44,7 +44,7 @@ class ConcatKernel : public framework::OpKernel<T> {
 };
 
 template <typename Place, typename T>
-class ConcatGradKernel : public framework::OpKernel {
+class ConcatGradKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const {
     auto* in = ctx.Input<framework::Tensor>(framework::GradVarName("Out"));
diff --git a/paddle/operators/reduce_op.h b/paddle/operators/reduce_op.h
index 2fbf94e34f3961a9b3140fb682a7c479f3b71f4d..ba3f3db81dc6251a063d27e597fd7e486e7b6c14 100644
--- a/paddle/operators/reduce_op.h
+++ b/paddle/operators/reduce_op.h
@@ -87,7 +87,7 @@ struct MaxOrMinGradFunctor {
 };
 
 template <typename Place, typename T, typename Functor>
-class ReduceKernel : public framework::OpKernel {
+class ReduceKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& context) const override {
     int rank = context.Input<Tensor>("X")->dims().size();
@@ -141,7 +141,7 @@ class ReduceKernel : public framework::OpKernel {
 };
 
 template <typename Place, typename T, typename Functor>
-class ReduceGradKernel : public framework::OpKernel {
+class ReduceGradKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& context) const override {
     int rank = context.Input<Tensor>("X")->dims().size();