Merge pull request #3859 from PaddlePaddle/revert-3824-remove_grad_op_type

Revert "Remove `grad_op_type` in `REGISTER_OP`"

doc/howto/dev/new_op_cn.md

@@ -178,13 +178,13 @@ class MulKernel : public framework::OpKernel {
 
 ```c++
 namespace ops = paddle::operators;
-REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, ops::MulOpGrad);
+REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulOpGrad);
 REGISTER_OP_CPU_KERNEL(mul, ops::MulKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(mul_grad,
               ops::MulGradKernel<paddle::platform::CPUPlace, float>);
 ```
     
-  - `REGISTER_OP` ： 注册`ops::MulOp`类，类型名为`mul`，该类的`ProtoMaker`为`ops::MulOpMaker`，并且注册`ops::MulOpGrad`为其反向Op。
+  - `REGISTER_OP` ： 注册`ops::MulOp`类，类型名为`mul`，该类的`ProtoMaker`为`ops::MulOpMaker`，注册`ops::MulOpGrad`，类型名为`mul_grad`，
   - `REGISTER_OP_WITHOUT_GRADIENT` ： 用于注册没有反向的Op。
   - `REGISTER_OP_CPU_KERNEL` ：注册`ops::MulKernel`类，并特化模板参数为`paddle::platform::CPUPlace`和`float`类型，同理，注册`ops::MulKernel`类。
 

paddle/framework/backward.md

@@ -18,7 +18,7 @@ A backward network is built up with several backward operators. Backward operato
 For example, we have got a `mul_op`, and we can register it's information and corresponding backward operator by the following macro:
 
 ```cpp
-REGISTER_OP(mul, MulOp, MulOpMaker, MulOpGrad);
+REGISTER_OP(mul, MulOp, MulOpMaker, mul_grad, MulOpGrad);
 ```
 
 `mul` is the operator's type. `MulOp` and `MulOpMaker` are the operator class and the operator maker class respectively.

paddle/framework/backward_test.cc

@@ -148,14 +148,16 @@ class AddOpMaker : public OpProtoAndCheckerMaker {
 namespace f = paddle::framework;
 namespace ops = paddle::operators;
 using EnforceNotMet = paddle::platform::EnforceNotMet;
-REGISTER_OP(rowwise_add, f::NOP, f::RowWiseAddOpMaker, f::NOP);
-REGISTER_OP(mul, f::NOP, f::MulOpMaker, f::NOP);
-REGISTER_OP(sigmoid, f::NOP, f::SigmoidOpMaker, f::NOP);
+REGISTER_OP(rowwise_add, f::NOP, f::RowWiseAddOpMaker, rowwise_add_grad,
+            f::NOP);
+REGISTER_OP(mul, f::NOP, f::MulOpMaker, mul_grad, f::NOP);
+REGISTER_OP(sigmoid, f::NOP, f::SigmoidOpMaker, sigmoid_grad, f::NOP);
 REGISTER_OP_WITHOUT_GRADIENT(nograd, f::NOP, f::NoGradOpMaker);
 REGISTER_OP_WITHOUT_GRADIENT(fill_zeros_like, f::NOP, f::FillZeroOpMaker);
-REGISTER_OP(add, f::NOP, f::AddOpMaker, f::NOP);
+REGISTER_OP(add, f::NOP, f::AddOpMaker, add_grad, f::NOP);
 REGISTER_OP_WITHOUT_GRADIENT(fc, f::FcOp, f::FcOpMaker);
-REGISTER_OP(many_output_op, f::NOP, f::ManyOutputOpMaker, f::NOP);
+REGISTER_OP(many_output_op, f::NOP, f::ManyOutputOpMaker, many_output_op_grad,
+            f::NOP);
 
 TEST(Backward, simple_op_grad) {
   auto fwd = f::OpRegistry::CreateOp(

paddle/framework/grad_op_builder_test.cc

@@ -54,8 +54,8 @@ TEST(GradOpBuilder, AddTwo) {
   EXPECT_EQ(grad_add_op->Output(f::GradVarName("Y")), f::GradVarName("y"));
 }
 
-REGISTER_OP(mult_io, f::NOP, f::MutiInOutOpMaker, f::NOP);
-REGISTER_OP(io_ignored, f::NOP, f::IOIgnoredOpMaker, f::NOP);
+REGISTER_OP(mult_io, f::NOP, f::MutiInOutOpMaker, mult_io_grad, f::NOP);
+REGISTER_OP(io_ignored, f::NOP, f::IOIgnoredOpMaker, io_ignored_grad, f::NOP);
 
 TEST(GradOpBuilder, MutiInOut) {
   std::shared_ptr<f::OperatorBase> test_op(f::OpRegistry::CreateOp(

paddle/framework/op_registry.h

@@ -33,7 +33,8 @@ namespace framework {
 class OpRegistry {
  public:
   template <typename OpType, typename ProtoMakerType, typename GradOpType>
-  static void RegisterOp(const std::string& op_type) {
+  static void RegisterOp(const std::string& op_type,
+                         const std::string& grad_op_type) {
     PADDLE_ENFORCE(!OpInfoMap::Instance().Has(op_type),
                    "'%s' is registered more than once.", op_type);
     OpInfo op_info;
@@ -42,9 +43,9 @@ class OpRegistry {
         const VariableNameMap& outputs, const AttributeMap& attrs) {
       return new OpType(type, inputs, outputs, attrs);
     };
+    op_info.grad_op_type_ = grad_op_type;
     if (std::type_index(typeid(ProtoMakerType)) !=
         std::type_index(typeid(NOPMaker))) {
-      op_info.grad_op_type_ = op_type + "_grad";
       op_info.proto_ = new OpProto;
       op_info.checker_ = new OpAttrChecker;
       auto maker = ProtoMakerType(op_info.proto_, op_info.checker_);
@@ -54,14 +55,15 @@ class OpRegistry {
           op_info.proto_->IsInitialized(),
           "Fail to initialize %s's OpProto, because %s is not initialized",
           op_type, op_info.proto_->InitializationErrorString());
-      // register gradient op
-      RegisterOp<GradOpType, NOPMaker, NOP>(op_info.grad_op_type_);
     } else {
-      op_info.grad_op_type_ = "";
       op_info.proto_ = nullptr;
       op_info.checker_ = nullptr;
     }
     OpInfoMap::Instance().Insert(op_type, op_info);
+    // register gradient op
+    if (!grad_op_type.empty()) {
+      RegisterOp<GradOpType, NOPMaker, NOP>(grad_op_type, "");
+    }
   }
 
   static std::unique_ptr<OperatorBase> CreateOp(const std::string& type,
@@ -90,8 +92,10 @@ class Registrar {
 template <typename OpType, typename ProtoMakerType, typename GradOpType>
 class OpRegistrar : public Registrar {
  public:
-  explicit OpRegistrar(const char* op_type) {
-    OpRegistry::RegisterOp<OpType, ProtoMakerType, GradOpType>(op_type);
+  explicit OpRegistrar(const char* op_type) { OpRegistrar(op_type, ""); }
+  OpRegistrar(const char* op_type, const char* grad_op_type) {
+    OpRegistry::RegisterOp<OpType, ProtoMakerType, GradOpType>(op_type,
+                                                               grad_op_type);
   }
 };
 
@@ -117,7 +121,8 @@ class OpKernelRegistrar : public Registrar {
 /**
  * Macro to register Operator.
  */
-#define REGISTER_OP(op_type, op_class, op_maker_class, grad_op_class)         \
+#define REGISTER_OP(op_type, op_class, op_maker_class, grad_op_type,          \
+                    grad_op_class)                                            \
   STATIC_ASSERT_GLOBAL_NAMESPACE(                                             \
       __reg_op__##op_type, "REGISTER_OP must be called in global namespace"); \
   class _OpClass_##op_type##_ : public op_class {                             \
@@ -132,14 +137,14 @@ class OpKernelRegistrar : public Registrar {
   };                                                                          \
   static ::paddle::framework::OpRegistrar<                                    \
       _OpClass_##op_type##_, op_maker_class, _OpGradClass_##op_type##_>       \
-      __op_registrar_##op_type##__(#op_type);                                 \
+      __op_registrar_##op_type##__(#op_type, #grad_op_type);                  \
   int TouchOpRegistrar_##op_type() {                                          \
     __op_registrar_##op_type##__.Touch();                                     \
     return 0;                                                                 \
   }
 
 #define REGISTER_OP_WITHOUT_GRADIENT(op_type, op_class, op_maker_class) \
-  REGISTER_OP(op_type, op_class, op_maker_class, ::paddle::framework::NOP)
+  REGISTER_OP(op_type, op_class, op_maker_class, , ::paddle::framework::NOP)
 
 /**
  * Macro to register OperatorKernel.

paddle/operators/add_op.cc

@@ -57,7 +57,7 @@ class AddOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(add_two, ops::AddOp, ops::AddOpMaker, ops::AddOpGrad);
+REGISTER_OP(add_two, ops::AddOp, ops::AddOpMaker, add_two_grad, ops::AddOpGrad);
 
 REGISTER_OP_CPU_KERNEL(add_two,
                        ops::AddKernel<paddle::platform::CPUPlace, float>);

paddle/operators/cross_entropy_op.cc

@@ -67,7 +67,8 @@ OnehotCrossEntropy Operator.
 
 namespace ops = paddle::operators;
 REGISTER_OP(onehot_cross_entropy, ops::OnehotCrossEntropyOp,
-            ops::OnehotCrossEntropyOpMaker, ops::OnehotCrossEntropyGradientOp);
+            ops::OnehotCrossEntropyOpMaker, onehot_cross_entropy_grad,
+            ops::OnehotCrossEntropyGradientOp);
 REGISTER_OP_CPU_KERNEL(onehot_cross_entropy,
                        ops::OnehotCrossEntropyOpKernel<float>);
 REGISTER_OP_CPU_KERNEL(onehot_cross_entropy_grad,

paddle/operators/gather_op.cc

@@ -63,7 +63,8 @@ Out = X[Index]
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(gather, ops::GatherOp, ops::GatherOpMaker, ops::GatherGradOp);
+REGISTER_OP(gather, ops::GatherOp, ops::GatherOpMaker, gather_grad,
+            ops::GatherGradOp);
 REGISTER_OP_CPU_KERNEL(gather,
                        ops::GatherOpKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/operators/lookup_table_op.cc

@@ -66,7 +66,7 @@ class LookupTableOpGrad : public framework::OperatorWithKernel {
 
 namespace ops = paddle::operators;
 REGISTER_OP(lookup_table, ops::LookupTableOp, ops::LookupTableOpMaker,
-            ops::LookupTableOpGrad);
+            lookup_table_grad, ops::LookupTableOpGrad);
 
 REGISTER_OP_CPU_KERNEL(lookup_table, ops::LookupTableKernel<float>);
 REGISTER_OP_CPU_KERNEL(lookup_table_grad, ops::LookupTableGradKernel<float>);

paddle/operators/mean_op.cc

@@ -54,7 +54,7 @@ class MeanGradOp : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(mean, ops::MeanOp, ops::MeanOpMaker, ops::MeanGradOp);
+REGISTER_OP(mean, ops::MeanOp, ops::MeanOpMaker, mean_grad, ops::MeanGradOp);
 REGISTER_OP_CPU_KERNEL(mean,
                        ops::MeanKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(mean_grad,

paddle/operators/minus_op.cc

@@ -81,6 +81,7 @@ class MinusGradOp : public NetOp {
 USE_OP(scale);
 USE_OP_ITSELF(identity);
 namespace ops = paddle::operators;
-REGISTER_OP(minus, ops::MinusOp, ops::MinusOpMaker, ops::MinusGradOp<float>);
+REGISTER_OP(minus, ops::MinusOp, ops::MinusOpMaker, minus_grad,
+            ops::MinusGradOp<float>);
 REGISTER_OP_CPU_KERNEL(minus,
                        ops::MinusKernel<paddle::platform::CPUPlace, float>);

paddle/operators/mul_op.cc

@@ -84,7 +84,7 @@ class MulOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, ops::MulOpGrad);
+REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulOpGrad);
 REGISTER_OP_CPU_KERNEL(mul, ops::MulKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(mul_grad,
                        ops::MulGradKernel<paddle::platform::CPUPlace, float>);

paddle/operators/rowwise_add_op.cc

@@ -74,7 +74,7 @@ class RowwiseAddGradOp : public framework::OperatorWithKernel {
 
 namespace ops = paddle::operators;
 REGISTER_OP(rowwise_add, ops::RowwiseAddOp, ops::RowwiseAddOpMaker,
-            ops::RowwiseAddGradOp);
+            rowwise_add_grad, ops::RowwiseAddGradOp);
 REGISTER_OP_CPU_KERNEL(
     rowwise_add, ops::RowwiseAddKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/operators/scale_op.cc

@@ -97,7 +97,7 @@ class IdentityOp : public NetOp {
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(scale, ops::ScaleOp, ops::ScaleOpMaker<float>,
+REGISTER_OP(scale, ops::ScaleOp, ops::ScaleOpMaker<float>, scale_grad,
             ops::ScaleGradOp<float>);
 REGISTER_OP_CPU_KERNEL(scale,
                        ops::ScaleKernel<paddle::platform::CPUPlace, float>);

paddle/operators/scatter_op.cc

@@ -77,7 +77,8 @@ Out[Index] = Ref[Index] + Updates
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(scatter, ops::ScatterOp, ops::ScatterOpMaker, ops::ScatterGradOp);
+REGISTER_OP(scatter, ops::ScatterOp, ops::ScatterOpMaker, scatter_grad,
+            ops::ScatterGradOp);
 REGISTER_OP_CPU_KERNEL(scatter,
                        ops::ScatterOpKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/operators/sigmoid_op.cc

@@ -53,7 +53,8 @@ class SigmoidOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP(sigmoid, ops::SigmoidOp, ops::SigmoidOpMaker, ops::SigmoidOpGrad);
+REGISTER_OP(sigmoid, ops::SigmoidOp, ops::SigmoidOpMaker, sigmoid_grad,
+            ops::SigmoidOpGrad);
 REGISTER_OP_CPU_KERNEL(sigmoid,
                        ops::SigmoidKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(

paddle/operators/softmax_op.cc

@@ -62,7 +62,8 @@ class SoftmaxOpGrad : public framework::OperatorWithKernel {
 
 namespace ops = paddle::operators;
 
-REGISTER_OP(softmax, ops::SoftmaxOp, ops::SoftmaxOpMaker, ops::SoftmaxOpGrad);
+REGISTER_OP(softmax, ops::SoftmaxOp, ops::SoftmaxOpMaker, softmax_grad,
+            ops::SoftmaxOpGrad);
 REGISTER_OP_CPU_KERNEL(softmax,
                        ops::SoftmaxKernel<paddle::platform::CPUPlace, float>);
 REGISTER_OP_CPU_KERNEL(