Add IndicateVarDataType interface to block tensor is not initialized problem...

Add IndicateVarDataType interface to block tensor is not initialized problem in OP GetExceptedKernelType (#20044)

* add indicate_var_data_type inferface, test=develop

* add unittests & polish error message, test=develop

* remove needless include, test=develop

* extract public function & polish message, test=develop

* delete empty var check, test=develop

* change data_type to pointer parameter, test=develop

* polish details, test=develop

paddle/fluid/framework/operator.cc

@@ -1138,40 +1138,65 @@ Scope* OperatorWithKernel::PrepareData(
   return new_scope;
 }
 
+void OperatorWithKernel::ParseInputDataType(
+    const ExecutionContext& ctx, const std::string& name,
+    proto::VarType::Type* data_type) const {
+  proto::VarType::Type dafault_data_type =
+      static_cast<proto::VarType::Type>(-1);
+  const std::vector<const Variable*> vars = ctx.MultiInputVar(name);
+  for (size_t i = 0; i < vars.size(); ++i) {
+    const Variable* var = vars[i];
+    if (var != nullptr) {
+      const Tensor* t = nullptr;
+      if (var->IsType<Tensor>()) {
+        t = &var->Get<Tensor>();
+      } else if (var->IsType<LoDTensor>()) {
+        t = &var->Get<LoDTensor>();
+      } else if (var->IsType<SelectedRows>()) {
+        t = &(var->Get<SelectedRows>().value());
+      }
+      if (t != nullptr) {
+        PADDLE_ENFORCE_EQ(t->IsInitialized(), true,
+                          "The Tensor in the %s Op's Input Variable %s(%s) is "
+                          "not initialized.",
+                          Type(), name, ctx.Inputs(name).at(i));
+        proto::VarType::Type tmp = t->type();
+        PADDLE_ENFORCE(tmp == *data_type || *data_type == dafault_data_type,
+                       "The DataType of %s Op's duplicable Variable %s must be "
+                       "consistent. The current variable type is (%s), but the "
+                       "previous variable type is (%s).",
+                       Type(), name, DataTypeToString(tmp),
+                       DataTypeToString(*data_type));
+        *data_type = tmp;
+      }
+    }
+  }
+}
+
 proto::VarType::Type OperatorWithKernel::IndicateDataType(
     const ExecutionContext& ctx) const {
   proto::VarType::Type dafault_data_type =
       static_cast<proto::VarType::Type>(-1);
   proto::VarType::Type data_type = dafault_data_type;
   for (auto& input : this->inputs_) {
-    const std::vector<const Variable*> vars = ctx.MultiInputVar(input.first);
-    for (size_t i = 0; i < vars.size(); ++i) {
-      const Variable* var = vars[i];
-      if (var != nullptr) {
-        const Tensor* t = nullptr;
-        if (var->IsType<Tensor>()) {
-          t = &var->Get<Tensor>();
-        } else if (var->IsType<LoDTensor>()) {
-          t = &var->Get<LoDTensor>();
-        } else if (var->IsType<SelectedRows>()) {
-          t = &(var->Get<SelectedRows>().value());
-        }
-        if (t != nullptr) {
-          PADDLE_ENFORCE(t->IsInitialized(), "Input %s(%lu) is not initialized",
-                         input.first, i);
-          proto::VarType::Type tmp = t->type();
-          PADDLE_ENFORCE(
-              tmp == data_type || data_type == dafault_data_type,
-              "DataType of Paddle Op %s %s must be the same. Get (%s) != (%s)",
-              Type(), input.first, DataTypeToString(data_type),
-              DataTypeToString(tmp));
-          data_type = tmp;
-        }
-      }
-    }
+    ParseInputDataType(ctx, input.first, &data_type);
   }
-  PADDLE_ENFORCE(data_type != dafault_data_type,
-                 "DataType should be indicated by input");
+  PADDLE_ENFORCE_NE(data_type, dafault_data_type,
+                    "DataType should be indicated by input Variable.");
+  return data_type;
+}
+
+proto::VarType::Type OperatorWithKernel::IndicateVarDataType(
+    const ExecutionContext& ctx, const std::string& name) const {
+  proto::VarType::Type dafault_data_type =
+      static_cast<proto::VarType::Type>(-1);
+  proto::VarType::Type data_type = dafault_data_type;
+  ParseInputDataType(ctx, name, &data_type);
+  PADDLE_ENFORCE_NE(
+      data_type, dafault_data_type,
+      "The Input Variable(%s) of %s Op used to determine kernel data type "
+      "is empty or not LoDTensor or SelectedRows.",
+      name, Type());
   return data_type;
 }
 

paddle/fluid/framework/operator.h

@@ -459,6 +459,9 @@ class OperatorWithKernel : public OperatorBase {
   void RuntimeInferShape(const Scope& scope, const platform::Place& place,
                          const RuntimeContext& ctx) const override;
 
+  proto::VarType::Type IndicateVarDataType(const ExecutionContext& ctx,
+                                           const std::string& name) const;
+
   virtual OpKernelType GetExpectedKernelType(const ExecutionContext& ctx) const;
 
   std::vector<KernelConfig>* GetKernelConfig(const OpKernelType& key) const;
@@ -470,6 +473,8 @@ class OperatorWithKernel : public OperatorBase {
       const OpKernelType& expected_kernel_type) const;
 
  private:
+  void ParseInputDataType(const ExecutionContext& ctx, const std::string& name,
+                          proto::VarType::Type* type) const;
   // indicate kernel DataType by input data. By default all input data must be
   // same.
   proto::VarType::Type IndicateDataType(const ExecutionContext& ctx) const;

paddle/fluid/framework/operator_test.cc

@@ -315,3 +315,182 @@ TEST(VarNameTest, all) {
   original_var_name = paddle::framework::GradOriginalVarName(original_var_name);
   ASSERT_EQ(original_var_name, "");
 }
+
+namespace paddle {
+namespace framework {
+
+class IndicateLoDTensorDataTypeTest : public OperatorWithKernel {
+ public:
+  using OperatorWithKernel::OperatorWithKernel;
+
+ protected:
+  void InferShape(framework::InferShapeContext* ctx) const override {}
+  OpKernelType GetExpectedKernelType(
+      const ExecutionContext& ctx) const override {
+    auto data_type = OperatorWithKernel::IndicateVarDataType(ctx, "LoDTensor");
+    return framework::OpKernelType(data_type, ctx.device_context());
+  }
+};
+class IndicateLoDTensorDataTypeTestProtoMaker : public OpProtoAndCheckerMaker {
+ public:
+  void Make() {
+    AddInput("LoDTensor", "Input of Tensor type Variable.");
+    AddComment("This Op is only for IndicateVarDataType inferface test.");
+  }
+};
+
+class IndicateSelectedRowsDataTypeTest : public OperatorWithKernel {
+ public:
+  using OperatorWithKernel::OperatorWithKernel;
+
+ protected:
+  void InferShape(framework::InferShapeContext* ctx) const override {}
+  OpKernelType GetExpectedKernelType(
+      const ExecutionContext& ctx) const override {
+    auto data_type =
+        OperatorWithKernel::IndicateVarDataType(ctx, "SelectedRows");
+    return framework::OpKernelType(data_type, ctx.device_context());
+  }
+};
+class IndicateSelectedRowsDataTypeTestProtoMaker
+    : public OpProtoAndCheckerMaker {
+ public:
+  void Make() {
+    AddInput("SelectedRows", "Input of SelectedRows type Variable.");
+    AddComment("This Op is only for IndicateVarDataType inferface test.");
+  }
+};
+
+class IndicateOtherDataTypeTest : public OperatorWithKernel {
+ public:
+  using OperatorWithKernel::OperatorWithKernel;
+
+ protected:
+  void InferShape(framework::InferShapeContext* ctx) const override {}
+  OpKernelType GetExpectedKernelType(
+      const ExecutionContext& ctx) const override {
+    auto data_type = OperatorWithKernel::IndicateVarDataType(ctx, "Other");
+    return framework::OpKernelType(data_type, ctx.device_context());
+  }
+};
+class IndicateOtherDataTypeTestProtoMaker : public OpProtoAndCheckerMaker {
+ public:
+  void Make() {
+    AddInput("Other", "Input of Other type Variable");
+    AddComment("This Op is only for IndicateVarDataType inferface test.");
+  }
+};
+
+template <typename DeviceContext, typename T>
+class IndicateVarDataTypeKernelTest : public OpKernel<T> {
+ public:
+  void Compute(const ExecutionContext& ctx) const {}
+};
+
+}  // namespace framework
+}  // namespace paddle
+
+REGISTER_OP_WITHOUT_GRADIENT(
+    indicate_lod_tensor_data_type_test,
+    paddle::framework::IndicateLoDTensorDataTypeTest,
+    paddle::framework::IndicateLoDTensorDataTypeTestProtoMaker);
+REGISTER_OP_WITHOUT_GRADIENT(
+    indicate_selected_rows_data_type_test,
+    paddle::framework::IndicateSelectedRowsDataTypeTest,
+    paddle::framework::IndicateSelectedRowsDataTypeTestProtoMaker);
+REGISTER_OP_WITHOUT_GRADIENT(
+    indicate_other_data_type_test, paddle::framework::IndicateOtherDataTypeTest,
+    paddle::framework::IndicateOtherDataTypeTestProtoMaker);
+
+REGISTER_OP_CPU_KERNEL(indicate_lod_tensor_data_type_test,
+                       paddle::framework::IndicateVarDataTypeKernelTest<
+                           paddle::platform::CPUDeviceContext, int>);
+REGISTER_OP_CPU_KERNEL(indicate_selected_rows_data_type_test,
+                       paddle::framework::IndicateVarDataTypeKernelTest<
+                           paddle::platform::CPUDeviceContext, int>);
+REGISTER_OP_CPU_KERNEL(indicate_other_data_type_test,
+                       paddle::framework::IndicateVarDataTypeKernelTest<
+                           paddle::platform::CPUDeviceContext, int>);
+
+TEST(IndicateVarDataTypeTest, lodtensor) {
+  paddle::framework::InitDevices(true);
+  paddle::framework::proto::OpDesc op_desc;
+  op_desc.set_type("indicate_lod_tensor_data_type_test");
+  BuildVar("LoDTensor", {"lodtensor_1"}, op_desc.add_inputs());
+
+  paddle::platform::CPUPlace cpu_place;
+  paddle::framework::Scope scope;
+
+  auto op = paddle::framework::OpRegistry::CreateOp(op_desc);
+  auto* var = scope.Var("lodtensor_1");
+  var->GetMutable<paddle::framework::LoDTensor>();
+
+  bool caught = false;
+  try {
+    op->Run(scope, cpu_place);
+  } catch (paddle::platform::EnforceNotMet err) {
+    caught = true;
+    std::string ex_msg = err.what();
+    EXPECT_TRUE(
+        ex_msg.find(
+            "The Tensor in the indicate_lod_tensor_data_type_test Op's "
+            "Input Variable LoDTensor(lodtensor_1) is not initialized") !=
+        std::string::npos);
+  }
+  ASSERT_TRUE(caught);
+}
+
+TEST(IndicateVarDataTypeTest, selectedrows) {
+  paddle::framework::InitDevices(true);
+  paddle::framework::proto::OpDesc op_desc;
+  op_desc.set_type("indicate_selected_rows_data_type_test");
+  BuildVar("SelectedRows", {"selected_rows_1"}, op_desc.add_inputs());
+
+  paddle::platform::CPUPlace cpu_place;
+  paddle::framework::Scope scope;
+
+  auto op = paddle::framework::OpRegistry::CreateOp(op_desc);
+  auto* var = scope.Var("selected_rows_1");
+  var->GetMutable<paddle::framework::SelectedRows>();
+
+  bool caught = false;
+  try {
+    op->Run(scope, cpu_place);
+  } catch (paddle::platform::EnforceNotMet err) {
+    caught = true;
+    std::string ex_msg = err.what();
+    EXPECT_TRUE(
+        ex_msg.find("The Tensor in the indicate_selected_rows_data_type_test "
+                    "Op's Input Variable SelectedRows(selected_rows_1) is not "
+                    "initialized") != std::string::npos);
+  }
+  ASSERT_TRUE(caught);
+}
+
+TEST(IndicateVarDataTypeTest, other) {
+  paddle::framework::InitDevices(true);
+  paddle::framework::proto::OpDesc op_desc;
+  op_desc.set_type("indicate_other_data_type_test");
+  BuildVar("Other", {"lod_tensor_array_1"}, op_desc.add_inputs());
+
+  paddle::platform::CPUPlace cpu_place;
+  paddle::framework::Scope scope;
+
+  auto op = paddle::framework::OpRegistry::CreateOp(op_desc);
+  auto* var = scope.Var("lod_tensor_array_1");
+  var->GetMutable<paddle::framework::LoDTensorArray>();
+
+  bool caught = false;
+  try {
+    op->Run(scope, cpu_place);
+  } catch (paddle::platform::EnforceNotMet err) {
+    caught = true;
+    std::string ex_msg = err.what();
+    EXPECT_TRUE(ex_msg.find("The Input Variable(Other) of "
+                            "indicate_other_data_type_test Op used to "
+                            "determine kernel data type "
+                            "is empty or not LoDTensor or SelectedRows") !=
+                std::string::npos);
+  }
+  ASSERT_TRUE(caught);
+}

paddle/fluid/framework/variable.h

@@ -30,9 +30,9 @@ class Variable {
     static_assert(
         IsRegisteredVarType<T>(),
         "Not registered type. Please register T inside var_type_traits.h");
-    PADDLE_ENFORCE(holder_ != nullptr, "Variable must hold some thing");
+    PADDLE_ENFORCE(holder_ != nullptr, "Variable is not initialized.");
     PADDLE_ENFORCE(holder_->Type() == VarTypeTrait<T>::kId,
-                   "Variable must be type %s, the holding type is %s",
+                   "The Variable type must be %s, but the type it holds is %s.",
                    ToTypeName(VarTypeTrait<T>::kId),
                    ToTypeName(holder_->Type()));
     return *static_cast<const T*>(holder_->Ptr());
@@ -45,10 +45,10 @@ class Variable {
     if (!holder_) {
       holder_.reset(new PlaceholderImpl<T>());
     } else {
-      PADDLE_ENFORCE(holder_->Type() == VarTypeTrait<T>::kId,
-                     "Variable must be type %s, the holding type is %s",
-                     ToTypeName(VarTypeTrait<T>::kId),
-                     ToTypeName(holder_->Type()));
+      PADDLE_ENFORCE(
+          holder_->Type() == VarTypeTrait<T>::kId,
+          "The Variable type must be %s, but the type it holds is %s.",
+          ToTypeName(VarTypeTrait<T>::kId), ToTypeName(holder_->Type()));
     }
     return static_cast<T*>(holder_->Ptr());
   }
@@ -61,7 +61,7 @@ class Variable {
   void Clear() { holder_.reset(); }
 
   int Type() const {
-    PADDLE_ENFORCE(holder_ != nullptr, "Must hold memory");
+    PADDLE_ENFORCE(holder_ != nullptr, "Variable is not initialized.");
     return holder_->Type();
   }
 

paddle/fluid/operators/mul_op.cc

@@ -92,7 +92,7 @@ class MulOp : public framework::OperatorWithKernel {
     framework::DataLayout layout = framework::DataLayout::kAnyLayout;
     int customized_type_value =
         framework::OpKernelType::kDefaultCustomizedTypeValue;
-    auto input_data_type = ctx.Input<Tensor>("X")->type();
+    auto input_data_type = OperatorWithKernel::IndicateVarDataType(ctx, "X");
 #ifdef PADDLE_WITH_MKLDNN
     if (library == framework::LibraryType::kPlain &&
         platform::CanMKLDNNBeUsed(ctx)) {

paddle/fluid/operators/save_op.cc

@@ -31,7 +31,7 @@ class SaveOp : public framework::OperatorWithKernel {
  protected:
   framework::OpKernelType GetExpectedKernelType(
       const framework::ExecutionContext &ctx) const override {
-    auto data_type = framework::GetDataTypeOfVar(ctx.InputVar("X"));
+    auto data_type = OperatorWithKernel::IndicateVarDataType(ctx, "X");
     return framework::OpKernelType(data_type, ctx.device_context());
   }
 };