diff --git a/paddle/operators/rowwise_add_op.cc b/paddle/operators/rowwise_add_op.cc
index cc763a8cf4565f633ee31b98c2ec7cd583ad77ed..178ea3c6145e00979b4eed1de99e81d1dd587fb4 100644
--- a/paddle/operators/rowwise_add_op.cc
+++ b/paddle/operators/rowwise_add_op.cc
@@ -16,7 +16,7 @@
 namespace paddle {
 namespace operators {
 
-class RowWiseAddOp : public OperatorWithKernel {
+class RowwiseAddOp : public OperatorWithKernel {
 protected:
   void InferShape(const InferShapeContext &ctx) const override {
     PADDLE_ENFORCE(ctx.InputSize() == 2UL,
@@ -32,9 +32,9 @@ protected:
   }
 };
 
-class RowWiseAddOpMaker : public OpProtoAndCheckerMaker {
+class RowwiseAddOpMaker : public OpProtoAndCheckerMaker {
 public:
-  RowWiseAddOpMaker(OpProto *proto, OpAttrChecker *op_checker)
+  RowwiseAddOpMaker(OpProto *proto, OpAttrChecker *op_checker)
       : OpProtoAndCheckerMaker(proto, op_checker) {
     AddInput("X", "The left input of row-wise add op, must be matrix");
     AddInput("b", "The right input of row-wise add op, must be vector");
@@ -46,13 +46,13 @@ for i in xrange(X.shape[0]):
 )DOC");
   }
 };
-class RowWiseAddGradOp : public OperatorWithKernel {
+class RowwiseAddGradOp : public OperatorWithKernel {
 protected:
   void InferShape(const InferShapeContext &ctx) const override {
     PADDLE_ENFORCE(ctx.InputSize() == 4UL,
-                   "RowWiseAddGrad inputs is I, O, OG, size must be 4");
+                   "RowwiseAddGrad inputs is I, O, OG, size must be 4");
     PADDLE_ENFORCE(ctx.OutputSize() == 2,
-                   "RowWiseAddGrad output is IG, size must be 2");
+                   "RowwiseAddGrad output is IG, size must be 2");
     ctx.Output<Tensor>(0)->Resize(ctx.Input<Tensor>(0)->dims());
     ctx.Output<Tensor>(1)->Resize(ctx.Input<Tensor>(1)->dims());
   }
@@ -61,10 +61,10 @@ protected:
 }  // namespace operators
 }  // namespace paddle
 
-REGISTER_OP(rowwise_add, ops::RowWiseAddOp, ops::RowWiseAddOpMaker);
+REGISTER_OP(rowwise_add, ops::RowwiseAddOp, ops::RowwiseAddOpMaker);
 REGISTER_OP_CPU_KERNEL(rowwise_add,
-                       ops::RowWiseAddKernel<ops::CPUPlace, float>);
+                       ops::RowwiseAddKernel<ops::CPUPlace, float>);
 
-REGISTER_GRADIENT_OP(rowwise_add, rowwise_add_grad, ops::RowWiseAddGradOp);
+REGISTER_GRADIENT_OP(rowwise_add, rowwise_add_grad, ops::RowwiseAddGradOp);
 REGISTER_OP_CPU_KERNEL(rowwise_add_grad,
-                       ops::RowWiseAddGradKernel<ops::CPUPlace, float>);
+                       ops::RowwiseAddGradKernel<ops::CPUPlace, float>);
diff --git a/paddle/operators/rowwise_add_op.cu b/paddle/operators/rowwise_add_op.cu
index 4b33e38ebabe853e179fe70ef7fde0a80b9050e2..f48dfeb6f2c516d8c1096885ad60dc333def6b1f 100644
--- a/paddle/operators/rowwise_add_op.cu
+++ b/paddle/operators/rowwise_add_op.cu
@@ -1,4 +1,4 @@
 #include "paddle/operators/rowwise_add_op.h"
 
 REGISTER_OP_GPU_KERNEL(rowwise_add,
-                       ops::RowWiseAddKernel<ops::GPUPlace, float>);
+                       ops::RowwiseAddKernel<ops::GPUPlace, float>);
diff --git a/paddle/operators/rowwise_add_op.h b/paddle/operators/rowwise_add_op.h
index 940459e0f176f07e63aef828baf3f9aa599c9a2c..321f51e61d472ede6cfc923fcf2a3d45324abd23 100644
--- a/paddle/operators/rowwise_add_op.h
+++ b/paddle/operators/rowwise_add_op.h
@@ -19,7 +19,7 @@ namespace paddle {
 namespace operators {
 
 template <typename Place, typename T>
-class RowWiseAddKernel : public OpKernel {
+class RowwiseAddKernel : public OpKernel {
 public:
   void Compute(const ExecutionContext& context) const override {
     auto out = context.Output<Tensor>(0);
@@ -39,7 +39,7 @@ public:
 };
 
 template <typename Place, typename T>
-class RowWiseAddGradKernel : public OpKernel {
+class RowwiseAddGradKernel : public OpKernel {
 public:
   void Compute(const ExecutionContext& context) const override {
     auto XGrad = context.Output<Tensor>(0);
@@ -51,7 +51,7 @@ public:
     auto OutGrad = EigenMatrix<T>::From(*context.Input<Tensor>(3));
     EigenMatrix<T>::From(*XGrad).device(*(context.GetEigenDevice<Place>())) =
         OutGrad;
-    // const int dimension = bGrad.dimension(0);
+
     // https://eigen.tuxfamily.org/dox/unsupported/TensorBase_8h_source.html
     EigenVector<T>::Flatten(*bGrad).device(*(context.GetEigenDevice<Place>())) =
         OutGrad.cumsum(1);  // colwise add