diff --git a/onnxruntime/python/onnxruntime_pybind_mlvalue.h b/onnxruntime/python/onnxruntime_pybind_mlvalue.h
index 9025b1eab4..85f8777aa3 100644
--- a/onnxruntime/python/onnxruntime_pybind_mlvalue.h
+++ b/onnxruntime/python/onnxruntime_pybind_mlvalue.h
@@ -28,6 +28,8 @@ MLDataType NumpyTypeToOnnxRuntimeType(int numpy_type);
 void CreateGenericMLValue(const onnxruntime::InputDefList* input_def_list, const AllocatorPtr& alloc, const std::string& name_input,
                           py::object& value, OrtValue* p_mlvalue);
 
+void GetPyObjFromTensor(const Tensor& rtensor, py::object& obj, const DataTransferManager* data_transfer_manager = nullptr);
+
 template <class T>
 struct DecRefFn {
   void operator()(T* pyobject) const {
diff --git a/onnxruntime/python/onnxruntime_pybind_state.cc b/onnxruntime/python/onnxruntime_pybind_state.cc
index 1724489159..f29e740820 100644
--- a/onnxruntime/python/onnxruntime_pybind_state.cc
+++ b/onnxruntime/python/onnxruntime_pybind_state.cc
@@ -189,7 +189,7 @@ void AddNonTensor(const OrtValue& val, std::vector<py::object>& pyobjs) {
   pyobjs.push_back(py::cast(val.Get<T>()));
 }
 
-void GetPyObjFromTensor(const Tensor& rtensor, py::object& obj, const DataTransferManager* data_transfer_manager = nullptr) {
+void GetPyObjFromTensor(const Tensor& rtensor, py::object& obj, const DataTransferManager* data_transfer_manager) {
   std::vector<npy_intp> npy_dims;
   const TensorShape& shape = rtensor.Shape();
 
diff --git a/onnxruntime/python/onnxruntime_pybind_state_common.h b/onnxruntime/python/onnxruntime_pybind_state_common.h
index 3855b480a8..b1b7637f2b 100644
--- a/onnxruntime/python/onnxruntime_pybind_state_common.h
+++ b/onnxruntime/python/onnxruntime_pybind_state_common.h
@@ -7,7 +7,10 @@
 #include "core/framework/session_options.h"
 
 #include "core/session/environment.h"
+
 namespace onnxruntime {
+class InferenceSession;
+
 namespace python {
 
 using namespace onnxruntime;
@@ -46,6 +49,8 @@ class SessionObjectInitializer {
 
 Environment& get_env();
 
+void InitializeSession(InferenceSession* sess, const std::vector<std::string>& provider_types);
+
 }
 }
 
diff --git a/orttraining/orttraining/python/orttraining_pybind_state.cc b/orttraining/orttraining/python/orttraining_pybind_state.cc
index 0885cd058b..a9d52c7a80 100644
--- a/orttraining/orttraining/python/orttraining_pybind_state.cc
+++ b/orttraining/orttraining/python/orttraining_pybind_state.cc
@@ -12,6 +12,8 @@
 #include "orttraining/core/session/training_session.h"
 #include "orttraining/core/graph/optimizer_config.h"
 #include "orttraining/core/framework/mpi_setup.h"
+#include "python/onnxruntime_pybind_mlvalue.h"
+
 
 namespace onnxruntime {
 namespace python {
@@ -19,13 +21,6 @@ namespace py = pybind11;
 using namespace onnxruntime;
 using namespace onnxruntime::logging;
 
-// BEGIN: forward declaration for stuff in onnxruntime_pybind_state
-void InitializeSession(InferenceSession* sess, const std::vector<std::string>& provider_types);
-void GetPyObjFromTensor(const Tensor& rtensor, py::object& obj, const DataTransferManager* data_transfer_manager = nullptr);
-void CreateGenericMLValue(const onnxruntime::InputDefList* input_def_list, const AllocatorPtr& alloc, const std::string& name_input,
-                          py::object& value, OrtValue* p_mlvalue);
-// END: forward declaration
-
 struct TrainingParameters {
   std::string loss_output_name;
   std::unordered_set<std::string> weights_to_train;