script to fix all

paddle/fluid/operators/activation_op.cc

@@ -558,95 +558,126 @@ $$out = \frac{x}{1 + e^{- \beta x}}$$
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(sigmoid, ops::ActivationOp, ops::SigmoidOpMaker, sigmoid_grad,
-            ops::ActivationOpGrad);
+REGISTER_OPERATOR(sigmoid, ops::ActivationOp, ops::SigmoidOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sigmoid_grad, ops::ActivationOpGrad)
 
-REGISTER_OP(logsigmoid, ops::ActivationOp, ops::LogSigmoidOpMaker,
-            logsigmoid_grad, ops::ActivationOpGrad);
+REGISTER_OPERATOR(logsigmoid, ops::ActivationOp, ops::LogSigmoidOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(logsigmoid_grad, ops::ActivationOpGrad)
 
-REGISTER_OP(exp, ops::ActivationOp, ops::ExpOpMaker, exp_grad,
-            ops::ActivationOpGrad);
+REGISTER_OPERATOR(exp, ops::ActivationOp, ops::ExpOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(exp_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(relu, ops::ActivationWithMKLDNNOp, ops::ReluOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(relu_grad, ops::ActivationWithMKLDNNOpGrad)
+
+REGISTER_OPERATOR(tanh, ops::ActivationWithMKLDNNOp, ops::TanhOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(tanh_grad, ops::ActivationWithMKLDNNOpGrad)
+
+REGISTER_OPERATOR(tanh_shrink, ops::ActivationOp, ops::TanhShrinkOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(tanh_shrink_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(softshrink, ops::ActivationOp, ops::SoftShrinkOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(softshrink_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(sqrt, ops::ActivationWithMKLDNNOp, ops::SqrtOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sqrt_grad, ops::ActivationWithMKLDNNOpGrad)
+
+REGISTER_OPERATOR(abs, ops::ActivationWithMKLDNNOp, ops::AbsOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(abs_grad, ops::ActivationWithMKLDNNOpGrad)
+
+REGISTER_OPERATOR(ceil, ops::ActivationOp, ops::CeilOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(ceil_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(floor, ops::ActivationOp, ops::FloorOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(floor_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(cos, ops::ActivationOp, ops::CosOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(cos_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(sin, ops::ActivationOp, ops::SinOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sin_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(round, ops::ActivationOp, ops::RoundOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(round_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(reciprocal, ops::ActivationOp, ops::ReciprocalOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reciprocal_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(log, ops::ActivationOp, ops::LogOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(log_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(square, ops::ActivationOp, ops::SquareOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(square_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(softplus, ops::ActivationOp, ops::SoftplusOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(softplus_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(softsign, ops::ActivationOp, ops::SoftsignOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(softsign_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(brelu, ops::ActivationOp, ops::BReluOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(brelu_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(leaky_relu, ops::ActivationOp, ops::LeakyReluOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(leaky_relu_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(soft_relu, ops::ActivationOp, ops::SoftReluOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(soft_relu_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(elu, ops::ActivationOp, ops::ELUOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(elu_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(relu6, ops::ActivationOp, ops::Relu6OpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(relu6_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(pow, ops::ActivationOp, ops::PowOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(pow_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(stanh, ops::ActivationOp, ops::STanhOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(stanh_grad, ops::ActivationOpGrad)
 
-REGISTER_OP(relu, ops::ActivationWithMKLDNNOp, ops::ReluOpMaker, relu_grad,
-            ops::ActivationWithMKLDNNOpGrad);
+REGISTER_OPERATOR(hard_shrink, ops::ActivationOp, ops::HardShrinkOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(hard_shrink_grad, ops::ActivationOpGrad)
 
-REGISTER_OP(tanh, ops::ActivationWithMKLDNNOp, ops::TanhOpMaker, tanh_grad,
-            ops::ActivationWithMKLDNNOpGrad);
-
-REGISTER_OP(tanh_shrink, ops::ActivationOp, ops::TanhShrinkOpMaker,
-            tanh_shrink_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(softshrink, ops::ActivationOp, ops::SoftShrinkOpMaker,
-            softshrink_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(sqrt, ops::ActivationWithMKLDNNOp, ops::SqrtOpMaker, sqrt_grad,
-            ops::ActivationWithMKLDNNOpGrad);
-
-REGISTER_OP(abs, ops::ActivationWithMKLDNNOp, ops::AbsOpMaker, abs_grad,
-            ops::ActivationWithMKLDNNOpGrad);
-
-REGISTER_OP(ceil, ops::ActivationOp, ops::CeilOpMaker, ceil_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(floor, ops::ActivationOp, ops::FloorOpMaker, floor_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(cos, ops::ActivationOp, ops::CosOpMaker, cos_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(sin, ops::ActivationOp, ops::SinOpMaker, sin_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(round, ops::ActivationOp, ops::RoundOpMaker, round_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(reciprocal, ops::ActivationOp, ops::ReciprocalOpMaker,
-            reciprocal_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(log, ops::ActivationOp, ops::LogOpMaker, log_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(square, ops::ActivationOp, ops::SquareOpMaker, square_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(softplus, ops::ActivationOp, ops::SoftplusOpMaker, softplus_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(softsign, ops::ActivationOp, ops::SoftsignOpMaker, softsign_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(brelu, ops::ActivationOp, ops::BReluOpMaker, brelu_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(leaky_relu, ops::ActivationOp, ops::LeakyReluOpMaker,
-            leaky_relu_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(soft_relu, ops::ActivationOp, ops::SoftReluOpMaker, soft_relu_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(elu, ops::ActivationOp, ops::ELUOpMaker, elu_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(relu6, ops::ActivationOp, ops::Relu6OpMaker, relu6_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(pow, ops::ActivationOp, ops::PowOpMaker, pow_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(stanh, ops::ActivationOp, ops::STanhOpMaker, stanh_grad,
-            ops::ActivationOpGrad);
-
-REGISTER_OP(hard_shrink, ops::ActivationOp, ops::HardShrinkOpMaker,
-            hard_shrink_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(thresholded_relu, ops::ActivationOp, ops::ThresholdedReluOpMaker,
-            thresholded_relu_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(hard_sigmoid, ops::ActivationOp, ops::HardSigmoidOpMaker,
-            hard_sigmoid_grad, ops::ActivationOpGrad);
-
-REGISTER_OP(swish, ops::ActivationOp, ops::SwishOpMaker, swish_grad,
-            ops::ActivationOpGrad);
+REGISTER_OPERATOR(thresholded_relu, ops::ActivationOp,
+                  ops::ThresholdedReluOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(thresholded_relu_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(hard_sigmoid, ops::ActivationOp, ops::HardSigmoidOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(hard_sigmoid_grad, ops::ActivationOpGrad)
+
+REGISTER_OPERATOR(swish, ops::ActivationOp, ops::SwishOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(swish_grad, ops::ActivationOpGrad)
 
 #define REGISTER_ACTIVATION_CPU_KERNEL(act_type, functor, grad_functor)   \
   REGISTER_OP_CPU_KERNEL(                                                 \

paddle/fluid/operators/bilinear_tensor_product_op.cc

@@ -153,9 +153,11 @@ class BilinearTensorProductOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(bilinear_tensor_product, ops::BilinearTensorProductOp,
-            ops::BilinearTensorProductOpMaker, bilinear_tensor_product_grad,
-            ops::BilinearTensorProductOpGrad);
+REGISTER_OPERATOR(bilinear_tensor_product, ops::BilinearTensorProductOp,
+                  ops::BilinearTensorProductOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(bilinear_tensor_product_grad,
+                  ops::BilinearTensorProductOpGrad)
 REGISTER_OP_CPU_KERNEL(
     bilinear_tensor_product,
     ops::BilinearTensorProductKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/clip_op.cc

@@ -81,8 +81,9 @@ class ClipOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(clip, ops::ClipOp, ops::ClipOpMaker<float>, clip_grad,
-            ops::ClipOpGrad);
+REGISTER_OPERATOR(clip, ops::ClipOp, ops::ClipOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(clip_grad, ops::ClipOpGrad)
 REGISTER_OP_CPU_KERNEL(
     clip, ops::ClipKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/conv_op.cc

@@ -335,14 +335,17 @@ framework::OpKernelType ConvOpGrad::GetExpectedKernelType(
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(conv2d, ops::ConvOp, ops::Conv2DOpMaker, conv2d_grad,
-            ops::ConvOpGrad);
+REGISTER_OPERATOR(conv2d, ops::ConvOp, ops::Conv2DOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(conv2d_grad, ops::ConvOpGrad)
 
 // depthwise convolution op
-REGISTER_OP(depthwise_conv2d, ops::ConvOp, ops::Conv2DOpMaker,
-            depthwise_conv2d_grad, ops::ConvOpGrad);
-REGISTER_OP(conv3d, ops::ConvOp, ops::Conv3DOpMaker, conv3d_grad,
-            ops::ConvOpGrad);
+REGISTER_OPERATOR(depthwise_conv2d, ops::ConvOp, ops::Conv2DOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(depthwise_conv2d_grad, ops::ConvOpGrad)
+REGISTER_OPERATOR(conv3d, ops::ConvOp, ops::Conv3DOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(conv3d_grad, ops::ConvOpGrad)
 
 // depthwise conv kernel
 // TODO(xingzhaolong): neon kernel for mobile

paddle/fluid/operators/conv_shift_op.cc

@@ -193,8 +193,9 @@ class ConvShiftGradKernel<platform::CPUPlace, T>
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(conv_shift, ops::ConvShiftOp, ops::ConvShiftOpMaker,
-            conv_shift_grad, ops::ConvShiftGradOp);
+REGISTER_OPERATOR(conv_shift, ops::ConvShiftOp, ops::ConvShiftOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(conv_shift_grad, ops::ConvShiftGradOp)
 REGISTER_OP_CPU_KERNEL(conv_shift,
                        ops::ConvShiftKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/conv_transpose_op.cc

@@ -298,8 +298,10 @@ framework::OpKernelType ConvTransposeOpGrad::GetExpectedKernelType(
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(conv2d_transpose, ops::ConvTransposeOp, ops::Conv2DTransposeOpMaker,
-            conv2d_transpose_grad, ops::ConvTransposeOpGrad);
+REGISTER_OPERATOR(conv2d_transpose, ops::ConvTransposeOp,
+                  ops::Conv2DTransposeOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(conv2d_transpose_grad, ops::ConvTransposeOpGrad)
 
 REGISTER_OP_CPU_KERNEL(
     conv2d_transpose,
@@ -311,8 +313,10 @@ REGISTER_OP_CPU_KERNEL(
     ops::GemmConvTransposeGradKernel<paddle::platform::CPUDeviceContext,
                                      double>);
 
-REGISTER_OP(conv3d_transpose, ops::ConvTransposeOp, ops::Conv3DTransposeOpMaker,
-            conv3d_transpose_grad, ops::ConvTransposeOpGrad);
+REGISTER_OPERATOR(conv3d_transpose, ops::ConvTransposeOp,
+                  ops::Conv3DTransposeOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(conv3d_transpose_grad, ops::ConvTransposeOpGrad)
 
 REGISTER_OP_CPU_KERNEL(
     conv3d_transpose,

paddle/fluid/operators/cos_sim_op.cc

@@ -153,8 +153,9 @@ class CosSimOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(cos_sim, ops::CosSimOp, ops::CosSimOpMaker, cos_sim_grad,
-            ops::CosSimOpGrad);
+REGISTER_OPERATOR(cos_sim, ops::CosSimOp, ops::CosSimOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(cos_sim_grad, ops::CosSimOpGrad)
 REGISTER_OP_CPU_KERNEL(
     cos_sim, ops::CosSimKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/cross_entropy_op.cc

@@ -164,8 +164,9 @@ or not. But the output only shares the LoD information with input X.
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(cross_entropy, ops::CrossEntropyOp, ops::CrossEntropyOpMaker,
-            cross_entropy_grad, ops::CrossEntropyGradientOp);
+REGISTER_OPERATOR(cross_entropy, ops::CrossEntropyOp, ops::CrossEntropyOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(cross_entropy_grad, ops::CrossEntropyGradientOp)
 REGISTER_OP_CPU_KERNEL(cross_entropy, ops::CrossEntropyOpKernel<float>,
                        ops::CrossEntropyOpKernel<double>);
 REGISTER_OP_CPU_KERNEL(cross_entropy_grad,

paddle/fluid/operators/dropout_op.cc

@@ -101,8 +101,9 @@ class DropoutOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(dropout, ops::DropoutOp, ops::DropoutOpMaker, dropout_grad,
-            ops::DropoutOpGrad);
+REGISTER_OPERATOR(dropout, ops::DropoutOp, ops::DropoutOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(dropout_grad, ops::DropoutOpGrad)
 REGISTER_OP_CPU_KERNEL(
     dropout, ops::CPUDropoutKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/elementwise_div_op.cc

@@ -30,8 +30,10 @@ class ElementwiseDivOpMaker : public ElementwiseOpMaker {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(elementwise_div, ops::ElementwiseOp, ops::ElementwiseDivOpMaker,
-            elementwise_div_grad, ops::ElementwiseOpGrad);
+REGISTER_OPERATOR(elementwise_div, ops::ElementwiseOp,
+                  ops::ElementwiseDivOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(elementwise_div_grad, ops::ElementwiseOpGrad)
 REGISTER_OP_CPU_KERNEL(
     elementwise_div,
     ops::ElementwiseDivKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/elementwise_max_op.cc

@@ -29,8 +29,10 @@ class ElementwiseMaxOpMaker : public ElementwiseOpMaker {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(elementwise_max, ops::ElementwiseOp, ops::ElementwiseMaxOpMaker,
-            elementwise_max_grad, ops::ElementwiseOpGrad);
+REGISTER_OPERATOR(elementwise_max, ops::ElementwiseOp,
+                  ops::ElementwiseMaxOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(elementwise_max_grad, ops::ElementwiseOpGrad)
 REGISTER_OP_CPU_KERNEL(
     elementwise_max,
     ops::ElementwiseMaxKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/elementwise_min_op.cc

@@ -29,8 +29,10 @@ class ElementwiseMinOpMaker : public ElementwiseOpMaker {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(elementwise_min, ops::ElementwiseOp, ops::ElementwiseMinOpMaker,
-            elementwise_min_grad, ops::ElementwiseOpGrad);
+REGISTER_OPERATOR(elementwise_min, ops::ElementwiseOp,
+                  ops::ElementwiseMinOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(elementwise_min_grad, ops::ElementwiseOpGrad)
 REGISTER_OP_CPU_KERNEL(
     elementwise_min,
     ops::ElementwiseMinKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/elementwise_mul_op.cc

@@ -31,8 +31,10 @@ class ElementwiseMulOpMaker : public ElementwiseOpMaker {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(elementwise_mul, ops::ElementwiseOp, ops::ElementwiseMulOpMaker,
-            elementwise_mul_grad, ops::ElementwiseOpGrad);
+REGISTER_OPERATOR(elementwise_mul, ops::ElementwiseOp,
+                  ops::ElementwiseMulOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(elementwise_mul_grad, ops::ElementwiseOpGrad)
 REGISTER_OP_CPU_KERNEL(
     elementwise_mul,
     ops::ElementwiseMulKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/elementwise_sub_op.cc

@@ -29,8 +29,10 @@ class ElementwiseSubOpMaker : public ElementwiseOpMaker {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(elementwise_sub, ops::ElementwiseOp, ops::ElementwiseSubOpMaker,
-            elementwise_sub_grad, ops::ElementwiseOpGrad);
+REGISTER_OPERATOR(elementwise_sub, ops::ElementwiseOp,
+                  ops::ElementwiseSubOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(elementwise_sub_grad, ops::ElementwiseOpGrad)
 REGISTER_OP_CPU_KERNEL(
     elementwise_sub,
     ops::ElementwiseSubKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/expand_op.cc

@@ -14,6 +14,8 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/expand_op.h"
 
+#include <vector>
+
 namespace paddle {
 namespace operators {
 
@@ -128,8 +130,9 @@ class ExpandGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(expand, ops::ExpandOp, ops::ExpandOpMaker, expand_grad,
-            ops::ExpandGradOp);
+REGISTER_OPERATOR(expand, ops::ExpandOp, ops::ExpandOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(expand_grad, ops::ExpandGradOp)
 REGISTER_OP_CPU_KERNEL(
     expand, ops::ExpandKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/fc_op.cc

@@ -98,5 +98,6 @@ FCOpMaker::FCOpMaker(OpProto* proto, OpAttrChecker* op_checker)
 }  // namespace operators
 }  // namespace paddle
 
-REGISTER_OP(fc, paddle::operators::FCOp, paddle::operators::FCOpMaker, fc_grad,
-            paddle::operators::FCOpGrad);
+REGISTER_OPERATOR(fc, paddle::operators::FCOp, paddle::operators::FCOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(fc_grad, paddle::operators::FCOpGrad)

paddle/fluid/operators/gather_op.cc

@@ -100,7 +100,8 @@ Out = [[3, 4],
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(gather, ops::GatherOp, ops::GatherOpMaker, gather_grad,
-            ops::GatherGradOp);
+REGISTER_OPERATOR(gather, ops::GatherOp, ops::GatherOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(gather_grad, ops::GatherGradOp)
 REGISTER_OP_CPU_KERNEL(gather, ops::GatherOpKernel<float>);
 REGISTER_OP_CPU_KERNEL(gather_grad, ops::GatherGradientOpKernel<float>);

paddle/fluid/operators/gru_op.cc

@@ -216,7 +216,9 @@ class GRUGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(gru, ops::GRUOp, ops::GRUOpMaker, gru_grad, ops::GRUGradOp);
+REGISTER_OPERATOR(gru, ops::GRUOp, ops::GRUOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(gru_grad, ops::GRUGradOp)
 REGISTER_OP_CPU_KERNEL(
     gru, ops::GRUKernel<paddle::platform::CPUDeviceContext, float>,
     ops::GRUKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/gru_unit_op.cc

@@ -198,8 +198,9 @@ class GRUUnitGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(gru_unit, ops::GRUUnitOp, ops::GRUUnitOpMaker, gru_unit_grad,
-            ops::GRUUnitGradOp);
+REGISTER_OPERATOR(gru_unit, ops::GRUUnitOp, ops::GRUUnitOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(gru_unit_grad, ops::GRUUnitGradOp)
 REGISTER_OP_CPU_KERNEL(
     gru_unit, ops::GRUUnitKernel<paddle::platform::CPUDeviceContext, float>,
     ops::GRUUnitKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/hinge_loss_op.cc

@@ -103,8 +103,9 @@ class HingeLossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(hinge_loss, ops::HingeLossOp, ops::HingeLossOpMaker<float>,
-            hinge_loss_grad, ops::HingeLossGradOp);
+REGISTER_OPERATOR(hinge_loss, ops::HingeLossOp, ops::HingeLossOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(hinge_loss_grad, ops::HingeLossGradOp)
 REGISTER_OP_CPU_KERNEL(
     hinge_loss,
     ops::HingeLossKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/huber_loss_op.cc

@@ -121,8 +121,9 @@ class HuberLossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(huber_loss, ops::HuberLossOp, ops::HuberLossOpMaker<float>,
-            huber_loss_grad, ops::HuberLossGradOp);
+REGISTER_OPERATOR(huber_loss, ops::HuberLossOp, ops::HuberLossOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(huber_loss_grad, ops::HuberLossGradOp)
 REGISTER_OP_CPU_KERNEL(
     huber_loss,
     ops::HuberLossKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/im2sequence_op.cc

@@ -148,8 +148,9 @@ class Im2SequenceGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(im2sequence, ops::Im2SequenceOp, ops::Im2SequenceOpMaker,
-            im2sequence_grad, ops::Im2SequenceGradOp);
+REGISTER_OPERATOR(im2sequence, ops::Im2SequenceOp, ops::Im2SequenceOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(im2sequence_grad, ops::Im2SequenceGradOp)
 REGISTER_OP_CPU_KERNEL(
     im2sequence,
     ops::Im2SequenceKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/l1_norm_op.cc

@@ -67,8 +67,9 @@ $$Out = \sum{|X|}$$
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(l1_norm, ops::L1NormOp, ops::L1NormOpMaker, l1_norm_grad,
-            ops::L1NormGradOp);
+REGISTER_OPERATOR(l1_norm, ops::L1NormOp, ops::L1NormOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(l1_norm_grad, ops::L1NormGradOp)
 REGISTER_OP_CPU_KERNEL(
     l1_norm, ops::L1NormKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/label_smooth_op.cc

@@ -117,8 +117,9 @@ class LabelSmoothGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 namespace ops = paddle::operators;
 
-REGISTER_OP(label_smooth, ops::LabelSmoothOp, ops::LabelSmoothOpMaker,
-            label_smooth_grad, ops::LabelSmoothGradOp);
+REGISTER_OPERATOR(label_smooth, ops::LabelSmoothOp, ops::LabelSmoothOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(label_smooth_grad, ops::LabelSmoothGradOp)
 REGISTER_OP_CPU_KERNEL(
     label_smooth,
     ops::LabelSmoothKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/layer_norm_op.cc

@@ -162,8 +162,9 @@ class LayerNormGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(layer_norm, ops::LayerNormOp, ops::LayerNormOpMaker,
-            layer_norm_grad, ops::LayerNormGradOp);
+REGISTER_OPERATOR(layer_norm, ops::LayerNormOp, ops::LayerNormOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(layer_norm_grad, ops::LayerNormGradOp)
 REGISTER_OP_CPU_KERNEL(
     layer_norm, ops::LayerNormKernel<paddle::platform::CPUDeviceContext, float>,
     ops::LayerNormKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/linear_chain_crf_op.cc

@@ -256,8 +256,10 @@ class LinearChainCRFGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(linear_chain_crf, ops::LinearChainCRFOp, ops::LinearChainCRFOpMaker,
-            linear_chain_crf_grad, ops::LinearChainCRFGradOp);
+REGISTER_OPERATOR(linear_chain_crf, ops::LinearChainCRFOp,
+                  ops::LinearChainCRFOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(linear_chain_crf_grad, ops::LinearChainCRFGradOp)
 REGISTER_OP_CPU_KERNEL(
     linear_chain_crf,
     ops::LinearChainCRFOpKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/lod_reset_op.cc

@@ -155,8 +155,9 @@ class LoDResetGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(lod_reset, ops::LoDResetOp, ops::LoDResetOpMaker, lod_reset_grad,
-            ops::LoDResetGradOp);
+REGISTER_OPERATOR(lod_reset, ops::LoDResetOp, ops::LoDResetOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(lod_reset_grad, ops::LoDResetGradOp)
 REGISTER_OP_CPU_KERNEL(
     lod_reset, ops::LoDResetKernel<paddle::platform::CPUPlace, float>,
     ops::LoDResetKernel<paddle::platform::CPUPlace, double>,

paddle/fluid/operators/log_loss_op.cc

@@ -106,8 +106,9 @@ class LogLossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(log_loss, ops::LogLossOp, ops::LogLossOpMaker<float>, log_loss_grad,
-            ops::LogLossGradOp);
+REGISTER_OPERATOR(log_loss, ops::LogLossOp, ops::LogLossOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(log_loss_grad, ops::LogLossGradOp)
 REGISTER_OP_CPU_KERNEL(
     log_loss, ops::LogLossKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/lrn_op.cc

@@ -276,7 +276,9 @@ class LRNOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(lrn, ops::LRNOp, ops::LRNOpMaker<float>, lrn_grad, ops::LRNOpGrad);
+REGISTER_OPERATOR(lrn, ops::LRNOp, ops::LRNOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(lrn_grad, ops::LRNOpGrad)
 REGISTER_OP_CPU_KERNEL(
     lrn, ops::LRNKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/lstm_op.cc

@@ -273,7 +273,9 @@ class LSTMGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(lstm, ops::LSTMOp, ops::LSTMOpMaker, lstm_grad, ops::LSTMGradOp);
+REGISTER_OPERATOR(lstm, ops::LSTMOp, ops::LSTMOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(lstm_grad, ops::LSTMGradOp)
 REGISTER_OP_CPU_KERNEL(
     lstm, ops::LSTMKernel<paddle::platform::CPUDeviceContext, float>,
     ops::LSTMKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/lstm_unit_op.cc

@@ -97,8 +97,9 @@ class LstmUnitGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(lstm_unit, ops::LstmUnitOp, ops::LstmUnitOpMaker, lstm_unit_grad,
-            ops::LstmUnitGradOp);
+REGISTER_OPERATOR(lstm_unit, ops::LstmUnitOp, ops::LstmUnitOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(lstm_unit_grad, ops::LstmUnitGradOp)
 REGISTER_OP_CPU_KERNEL(lstm_unit,
                        ops::LstmUnitKernel<paddle::platform::CPUPlace, float>,
                        ops::LstmUnitKernel<paddle::platform::CPUPlace, double>);

paddle/fluid/operators/lstmp_op.cc

@@ -322,8 +322,9 @@ class LSTMPGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(lstmp, ops::LSTMPOp, ops::LSTMPOpMaker, lstmp_grad,
-            ops::LSTMPGradOp);
+REGISTER_OPERATOR(lstmp, ops::LSTMPOp, ops::LSTMPOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(lstmp_grad, ops::LSTMPGradOp)
 REGISTER_OP_CPU_KERNEL(
     lstmp, ops::LSTMPKernel<paddle::platform::CPUDeviceContext, float>,
     ops::LSTMPKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/margin_rank_loss_op.cc

@@ -111,9 +111,10 @@ class MarginRankLossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 namespace ops = paddle::operators;
 
-REGISTER_OP(margin_rank_loss, ops::MarginRankLossOp,
-            ops::MarginRankLossOpMaker<float>, margin_rank_loss_grad,
-            ops::MarginRankLossGradOp);
+REGISTER_OPERATOR(margin_rank_loss, ops::MarginRankLossOp,
+                  ops::MarginRankLossOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(margin_rank_loss_grad, ops::MarginRankLossGradOp)
 REGISTER_OP_CPU_KERNEL(
     margin_rank_loss,
     ops::MarginRankLossKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/matmul_op.cc

@@ -237,8 +237,9 @@ class MatMulOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(matmul, ops::MatMulOp, ops::MatMulOpMaker, matmul_grad,
-            ops::MatMulOpGrad);
+REGISTER_OPERATOR(matmul, ops::MatMulOp, ops::MatMulOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(matmul_grad, ops::MatMulOpGrad)
 REGISTER_OP_CPU_KERNEL(
     matmul, ops::MatMulKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/maxout_op.cc

@@ -101,8 +101,9 @@ class MaxOutOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(maxout, ops::MaxOutOp, ops::MaxOutOpMaker, maxout_grad,
-            ops::MaxOutOpGrad);
+REGISTER_OPERATOR(maxout, ops::MaxOutOp, ops::MaxOutOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(maxout_grad, ops::MaxOutOpGrad)
 REGISTER_OP_CPU_KERNEL(
     maxout, ops::MaxOutKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/modified_huber_loss_op.cc

@@ -108,9 +108,10 @@ class ModifiedHuberLossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(modified_huber_loss, ops::ModifiedHuberLossOp,
-            ops::ModifiedHuberLossOpMaker, modified_huber_loss_grad,
-            ops::ModifiedHuberLossGradOp);
+REGISTER_OPERATOR(modified_huber_loss, ops::ModifiedHuberLossOp,
+                  ops::ModifiedHuberLossOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(modified_huber_loss_grad, ops::ModifiedHuberLossGradOp)
 
 REGISTER_OP_CPU_KERNEL(
     modified_huber_loss,

paddle/fluid/operators/mul_op.cc

@@ -160,7 +160,9 @@ class MulGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulGradOp);
+REGISTER_OPERATOR(mul, ops::MulOp, ops::MulOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(mul_grad, ops::MulGradOp)
 REGISTER_OP_CPU_KERNEL(
     mul, ops::MulKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/nce_op.cc

@@ -14,6 +14,8 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/nce_op.h"
 
+#include <vector>
+
 namespace paddle {
 namespace operators {
 
@@ -179,7 +181,9 @@ class NCEOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(nce, ops::NCEOp, ops::NCEOpMaker, nce_grad, ops::NCEOpGrad);
+REGISTER_OPERATOR(nce, ops::NCEOp, ops::NCEOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(nce_grad, ops::NCEOpGrad)
 REGISTER_OP_CPU_KERNEL(nce, ops::NCEKernel<paddle::platform::CPUPlace, float>,
                        ops::NCEKernel<paddle::platform::CPUPlace, double>);
 REGISTER_OP_CPU_KERNEL(nce_grad,

paddle/fluid/operators/norm_op.cc

@@ -85,8 +85,9 @@ class NormOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(norm, ops::NormOp, ops::NormOpMaker<float>, norm_grad,
-            ops::NormOpGrad);
+REGISTER_OPERATOR(norm, ops::NormOp, ops::NormOpMaker<float>,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(norm_grad, ops::NormOpGrad)
 REGISTER_OP_CPU_KERNEL(
     norm, ops::NormKernel<paddle::platform::CPUDeviceContext, float>,
     ops::NormKernel<paddle::platform::CPUDeviceContext, double, float>);

paddle/fluid/operators/pool_op.cc

@@ -333,8 +333,9 @@ Example:
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(pool2d, ops::PoolOp, ops::Pool2dOpMaker, pool2d_grad,
-            ops::PoolOpGrad);
+REGISTER_OPERATOR(pool2d, ops::PoolOp, ops::Pool2dOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(pool2d_grad, ops::PoolOpGrad)
 
 REGISTER_OP_CPU_KERNEL(
     pool2d, ops::PoolKernel<paddle::platform::CPUDeviceContext, float>,
@@ -343,8 +344,9 @@ REGISTER_OP_CPU_KERNEL(
     pool2d_grad, ops::PoolGradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::PoolGradKernel<paddle::platform::CPUDeviceContext, double>)
 
-REGISTER_OP(pool3d, ops::PoolOp, ops::Pool3dOpMaker, pool3d_grad,
-            ops::PoolOpGrad);
+REGISTER_OPERATOR(pool3d, ops::PoolOp, ops::Pool3dOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(pool3d_grad, ops::PoolOpGrad)
 
 REGISTER_OP_CPU_KERNEL(
     pool3d, ops::PoolKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/pool_with_index_op.cc

@@ -258,9 +258,10 @@ Example:
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(max_pool2d_with_index, ops::MaxPoolWithIndexOp,
-            ops::MaxPool2dWithIndexOpMaker, max_pool2d_with_index_grad,
-            ops::MaxPoolWithIndexOpGrad);
+REGISTER_OPERATOR(max_pool2d_with_index, ops::MaxPoolWithIndexOp,
+                  ops::MaxPool2dWithIndexOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(max_pool2d_with_index_grad, ops::MaxPoolWithIndexOpGrad)
 
 REGISTER_OP_CPU_KERNEL(
     max_pool2d_with_index,
@@ -274,9 +275,10 @@ REGISTER_OP_CPU_KERNEL(
     ops::MaxPoolWithIndexGradKernel<paddle::platform::CPUDeviceContext, double,
                                     int>)
 
-REGISTER_OP(max_pool3d_with_index, ops::MaxPoolWithIndexOp,
-            ops::MaxPool3dWithIndexOpMaker, max_pool3d_with_index_grad,
-            ops::MaxPoolWithIndexOpGrad);
+REGISTER_OPERATOR(max_pool3d_with_index, ops::MaxPoolWithIndexOp,
+                  ops::MaxPool3dWithIndexOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(max_pool3d_with_index_grad, ops::MaxPoolWithIndexOpGrad)
 
 REGISTER_OP_CPU_KERNEL(
     max_pool3d_with_index,

paddle/fluid/operators/prelu_op.cc

@@ -83,8 +83,9 @@ class PReluGradOp : public framework::OperatorWithKernel {
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(prelu, ops::PReluOp, ops::PReluOpMaker, prelu_grad,
-            ops::PReluGradOp);
+REGISTER_OPERATOR(prelu, ops::PReluOp, ops::PReluOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(prelu_grad, ops::PReluGradOp)
 REGISTER_OP_CPU_KERNEL(
     prelu, ops::PReluKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/rank_loss_op.cc

@@ -121,8 +121,9 @@ class RankLossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 namespace ops = paddle::operators;
 
-REGISTER_OP(rank_loss, ops::RankLossOp, ops::RankLossOpMaker, rank_loss_grad,
-            ops::RankLossGradOp);
+REGISTER_OPERATOR(rank_loss, ops::RankLossOp, ops::RankLossOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(rank_loss_grad, ops::RankLossGradOp)
 REGISTER_OP_CPU_KERNEL(
     rank_loss, ops::RankLossKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/reduce_op.cc

@@ -14,6 +14,9 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/reduce_op.h"
 
+#include <string>
+#include <vector>
+
 namespace paddle {
 namespace operators {
 
@@ -122,18 +125,18 @@ If reduce_all is true, just reduce along all dimensions and output a scalar.
  protected:
   std::string comment_;
 
-  void Replace(std::string &src, std::string from, std::string to) {
+  void Replace(std::string *src, std::string from, std::string to) {
     std::size_t len_from = std::strlen(from.c_str());
     std::size_t len_to = std::strlen(to.c_str());
-    for (std::size_t pos = src.find(from); pos != std::string::npos;
-         pos = src.find(from, pos + len_to)) {
-      src.replace(pos, len_from, to);
+    for (std::size_t pos = src->find(from); pos != std::string::npos;
+         pos = src->find(from, pos + len_to)) {
+      src->replace(pos, len_from, to);
     }
   }
 
   void SetComment(std::string name, std::string op) {
-    Replace(comment_, "{ReduceOp}", name);
-    Replace(comment_, "{reduce}", op);
+    Replace(&comment_, "{ReduceOp}", name);
+    Replace(&comment_, "{reduce}", op);
   }
 };
 
@@ -187,20 +190,25 @@ class ReduceProdOpMaker : public ReduceOpMaker {
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(reduce_sum, ops::ReduceOp, ops::ReduceSumOpMaker, reduce_sum_grad,
-            ops::ReduceGradOp);
+REGISTER_OPERATOR(reduce_sum, ops::ReduceOp, ops::ReduceSumOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reduce_sum_grad, ops::ReduceGradOp)
 
-REGISTER_OP(reduce_mean, ops::ReduceOp, ops::ReduceMeanOpMaker,
-            reduce_mean_grad, ops::ReduceGradOp);
+REGISTER_OPERATOR(reduce_mean, ops::ReduceOp, ops::ReduceMeanOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reduce_mean_grad, ops::ReduceGradOp)
 
-REGISTER_OP(reduce_max, ops::ReduceOp, ops::ReduceMaxOpMaker, reduce_max_grad,
-            ops::ReduceGradOp);
+REGISTER_OPERATOR(reduce_max, ops::ReduceOp, ops::ReduceMaxOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reduce_max_grad, ops::ReduceGradOp)
 
-REGISTER_OP(reduce_min, ops::ReduceOp, ops::ReduceMinOpMaker, reduce_min_grad,
-            ops::ReduceGradOp);
+REGISTER_OPERATOR(reduce_min, ops::ReduceOp, ops::ReduceMinOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reduce_min_grad, ops::ReduceGradOp)
 
-REGISTER_OP(reduce_prod, ops::ReduceOp, ops::ReduceProdOpMaker,
-            reduce_prod_grad, ops::ReduceGradOp);
+REGISTER_OPERATOR(reduce_prod, ops::ReduceOp, ops::ReduceProdOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reduce_prod_grad, ops::ReduceGradOp)
 
 #define REGISTER_REDUCE_CPU_KERNEL(reduce_type, functor, grad_functor)         \
   REGISTER_OP_CPU_KERNEL(reduce_type,                                          \

paddle/fluid/operators/reshape_op.cc

@@ -113,8 +113,9 @@ class ReshapeGradOp : public framework::OperatorWithKernel {
 namespace ops = paddle::operators;
 using CPU = paddle::platform::CPUDeviceContext;
 
-REGISTER_OP(reshape, ops::ReshapeOp, ops::ReshapeOpMaker, reshape_grad,
-            ops::ReshapeGradOp);
+REGISTER_OPERATOR(reshape, ops::ReshapeOp, ops::ReshapeOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(reshape_grad, ops::ReshapeGradOp)
 REGISTER_OP_CPU_KERNEL(reshape, ops::ReshapeKernel<CPU, float>,
                        ops::ReshapeKernel<CPU, double>,
                        ops::ReshapeKernel<CPU, int>,

paddle/fluid/operators/roi_pool_op.cc

@@ -153,8 +153,9 @@ https://stackoverflow.com/questions/43430056/what-is-roi-layer-in-fast-rcnn
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(roi_pool, ops::ROIPoolOp, ops::ROIPoolOpMaker, roi_pool_grad,
-            ops::ROIPoolGradOp);
+REGISTER_OPERATOR(roi_pool, ops::ROIPoolOp, ops::ROIPoolOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(roi_pool_grad, ops::ROIPoolGradOp)
 REGISTER_OP_CPU_KERNEL(
     roi_pool,
     ops::CPUROIPoolOpKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/row_conv_op.cc

@@ -250,8 +250,9 @@ class RowConvGradKernel<platform::CPUDeviceContext, T>
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(row_conv, ops::RowConvOp, ops::RowConvOpMaker, row_conv_grad,
-            ops::RowConvGradOp);
+REGISTER_OPERATOR(row_conv, ops::RowConvOp, ops::RowConvOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(row_conv_grad, ops::RowConvGradOp)
 REGISTER_OP_CPU_KERNEL(
     row_conv, ops::RowConvKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/scatter_op.cc

@@ -102,7 +102,8 @@ $$
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(scatter, ops::ScatterOp, ops::ScatterOpMaker, scatter_grad,
-            ops::ScatterGradOp);
+REGISTER_OPERATOR(scatter, ops::ScatterOp, ops::ScatterOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(scatter_grad, ops::ScatterGradOp)
 REGISTER_OP_CPU_KERNEL(scatter, ops::ScatterOpKernel<float>);
 REGISTER_OP_CPU_KERNEL(scatter_grad, ops::ScatterGradientOpKernel<float>);

paddle/fluid/operators/sequence_conv_op.cc

@@ -14,6 +14,8 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/sequence_conv_op.h"
 
+#include <algorithm>
+
 namespace paddle {
 namespace operators {
 
@@ -174,8 +176,9 @@ context_length, context_stride and context_start.
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(sequence_conv, ops::SequenceConvOp, ops::SequenceConvOpMaker,
-            sequence_conv_grad, ops::SequenceConvGradOp);
+REGISTER_OPERATOR(sequence_conv, ops::SequenceConvOp, ops::SequenceConvOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sequence_conv_grad, ops::SequenceConvGradOp)
 
 REGISTER_OP_CPU_KERNEL(
     sequence_conv,

paddle/fluid/operators/sequence_expand_op.cc

@@ -200,8 +200,10 @@ class SequenceExpandOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(sequence_expand, ops::SequenceExpandOp, ops::SequenceExpandOpMaker,
-            sequence_expand_grad, ops::SequenceExpandOpGrad);
+REGISTER_OPERATOR(sequence_expand, ops::SequenceExpandOp,
+                  ops::SequenceExpandOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sequence_expand_grad, ops::SequenceExpandOpGrad)
 REGISTER_OP_CPU_KERNEL(
     sequence_expand,
     ops::SequenceExpandKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/sequence_slice_op.cc

@@ -120,8 +120,10 @@ NOTE: The first dimension size of input, the size of offset and Length, should b
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(sequence_slice, ops::SequenceSliceOp, ops::SequenceSliceOpMaker,
-            sequence_slice_grad, ops::SequenceSliceGradOp);
+REGISTER_OPERATOR(sequence_slice, ops::SequenceSliceOp,
+                  ops::SequenceSliceOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sequence_slice_grad, ops::SequenceSliceGradOp)
 REGISTER_OP_CPU_KERNEL(
     sequence_slice,
     ops::SequenceSliceOpKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/sequence_softmax_op.cc

@@ -15,6 +15,8 @@ limitations under the License. */
 #include "paddle/fluid/operators/sequence_softmax_op.h"
 #include <string>
 
+#include <string>
+
 namespace paddle {
 namespace operators {
 
@@ -155,9 +157,10 @@ class SequenceSoftmaxGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(sequence_softmax, ops::SequenceSoftmaxOp,
-            ops::SequenceSoftmaxOpMaker, sequence_softmax_grad,
-            ops::SequenceSoftmaxGradOp);
+REGISTER_OPERATOR(sequence_softmax, ops::SequenceSoftmaxOp,
+                  ops::SequenceSoftmaxOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sequence_softmax_grad, ops::SequenceSoftmaxGradOp)
 REGISTER_OP_CPU_KERNEL(
     sequence_softmax,
     ops::SequenceSoftmaxKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/sigmoid_cross_entropy_with_logits_op.cc

@@ -135,11 +135,12 @@ However the output only shares the LoD with input `X`.
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(sigmoid_cross_entropy_with_logits,
-            ops::SigmoidCrossEntropyWithLogitsOp,
-            ops::SigmoidCrossEntropyWithLogitsOpMaker,
-            sigmoid_cross_entropy_with_logits_grad,
-            ops::SigmoidCrossEntropyWithLogitsGradOp);
+REGISTER_OPERATOR(sigmoid_cross_entropy_with_logits,
+                  ops::SigmoidCrossEntropyWithLogitsOp,
+                  ops::SigmoidCrossEntropyWithLogitsOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(sigmoid_cross_entropy_with_logits_grad,
+                  ops::SigmoidCrossEntropyWithLogitsGradOp)
 REGISTER_OP_CPU_KERNEL(sigmoid_cross_entropy_with_logits,
                        ops::SigmoidCrossEntropyWithLogitsKernel<
                            paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/smooth_l1_loss_op.cc

@@ -132,8 +132,9 @@ class SmoothL1LossGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(smooth_l1_loss, ops::SmoothL1LossOp, ops::SmoothL1LossOpMaker,
-            smooth_l1_loss_grad, ops::SmoothL1LossGradOp);
+REGISTER_OPERATOR(smooth_l1_loss, ops::SmoothL1LossOp, ops::SmoothL1LossOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(smooth_l1_loss_grad, ops::SmoothL1LossGradOp)
 REGISTER_OP_CPU_KERNEL(
     smooth_l1_loss,
     ops::SmoothL1LossKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/softmax_op.cc

@@ -157,8 +157,9 @@ class SoftmaxOpGrad : public framework::OperatorWithKernel {
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(softmax, ops::SoftmaxOp, ops::SoftmaxOpMaker, softmax_grad,
-            ops::SoftmaxOpGrad);
+REGISTER_OPERATOR(softmax, ops::SoftmaxOp, ops::SoftmaxOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(softmax_grad, ops::SoftmaxOpGrad)
 REGISTER_OP_CPU_KERNEL(
     softmax, ops::SoftmaxKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/spp_op.cc

@@ -92,7 +92,9 @@ class SppOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(spp, ops::SppOp, ops::SppOpMaker, spp_grad, ops::SppOpGrad);
+REGISTER_OPERATOR(spp, ops::SppOp, ops::SppOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(spp_grad, ops::SppOpGrad)
 REGISTER_OP_CPU_KERNEL(
     spp, ops::SppKernel<paddle::platform::CPUDeviceContext, float>,
     ops::SppKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/squared_l2_distance_op.cc

@@ -109,9 +109,10 @@ class SquaredL2DistanceGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(squared_l2_distance, ops::SquaredL2DistanceOp,
-            ops::SquaredL2DistanceOpMaker, squared_l2_distance_grad,
-            ops::SquaredL2DistanceGradOp);
+REGISTER_OPERATOR(squared_l2_distance, ops::SquaredL2DistanceOp,
+                  ops::SquaredL2DistanceOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(squared_l2_distance_grad, ops::SquaredL2DistanceGradOp)
 REGISTER_OP_CPU_KERNEL(
     squared_l2_distance,
     ops::SquaredL2DistanceKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/squared_l2_norm_op.cc

@@ -67,8 +67,10 @@ $$Out = \sum_{i} X_{i}^2$$
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(squared_l2_norm, ops::SquaredL2NormOp, ops::SquaredL2NormOpMaker,
-            squared_l2_norm_grad, ops::SquaredL2NormGradOp);
+REGISTER_OPERATOR(squared_l2_norm, ops::SquaredL2NormOp,
+                  ops::SquaredL2NormOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(squared_l2_norm_grad, ops::SquaredL2NormGradOp)
 REGISTER_OP_CPU_KERNEL(
     squared_l2_norm,
     ops::SquaredL2NormKernel<paddle::platform::CPUDeviceContext, float>);

paddle/fluid/operators/transpose_op.cc

@@ -118,8 +118,9 @@ class TransposeOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(transpose, ops::TransposeOp, ops::TransposeOpMaker, transpose_grad,
-            ops::TransposeOpGrad);
+REGISTER_OPERATOR(transpose, ops::TransposeOp, ops::TransposeOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(transpose_grad, ops::TransposeOpGrad)
 REGISTER_OP_CPU_KERNEL(
     transpose, ops::TransposeKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/unpool_op.cc

@@ -132,8 +132,9 @@ class UnpoolOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(unpool, ops::UnpoolOp, ops::Unpool2dOpMaker, unpool_grad,
-            ops::UnpoolOpGrad);
+REGISTER_OPERATOR(unpool, ops::UnpoolOp, ops::Unpool2dOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(unpool_grad, ops::UnpoolOpGrad)
 REGISTER_OP_CPU_KERNEL(
     unpool, ops::UnpoolKernel<paddle::platform::CPUDeviceContext, float>,
     ops::UnpoolKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/warpctc_op.cc

@@ -132,8 +132,9 @@ class WarpCTCGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(warpctc, ops::WarpCTCOp, ops::WarpCTCOpMaker, warpctc_grad,
-            ops::WarpCTCGradOp);
+REGISTER_OPERATOR(warpctc, ops::WarpCTCOp, ops::WarpCTCOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>)
+REGISTER_OPERATOR(warpctc_grad, ops::WarpCTCGradOp)
 REGISTER_OP_CPU_KERNEL(
     warpctc, ops::WarpCTCKernel<paddle::platform::CPUDeviceContext, float>);
 REGISTER_OP_CPU_KERNEL(