diff --git a/cmake/onnxruntime_unittests.cmake b/cmake/onnxruntime_unittests.cmake
index 66aef37825..8ac52f5ff3 100644
--- a/cmake/onnxruntime_unittests.cmake
+++ b/cmake/onnxruntime_unittests.cmake
@@ -91,6 +91,7 @@ file(GLOB onnxruntime_test_ir_src
 
 set(onnxruntime_test_framework_src_patterns
   "${TEST_SRC_DIR}/framework/*.cc"
+  "${TEST_SRC_DIR}/framework/*.h"
   "${TEST_SRC_DIR}/platform/*.cc"
   )
 
diff --git a/include/onnxruntime/core/framework/execution_provider.h b/include/onnxruntime/core/framework/execution_provider.h
index 8726dcf035..2f311161b8 100644
--- a/include/onnxruntime/core/framework/execution_provider.h
+++ b/include/onnxruntime/core/framework/execution_provider.h
@@ -38,6 +38,9 @@ struct NodeComputeInfo {
 };
 
 class IExecutionProvider {
+ protected:
+  IExecutionProvider(const std::string& type) : type_{type} {}
+
  public:
   virtual ~IExecutionProvider() = default;
 
@@ -105,7 +108,7 @@ class IExecutionProvider {
      through the SetExecutionProvider API. Example valid return values are:
      kCpuExecutionProvider, kCudaExecutionProvider
   */
-  virtual std::string Type() const = 0;
+  const std::string& Type() const { return type_; }
 
   /**
      Blocks until the device has completed all preceding requested tasks.
@@ -149,6 +152,7 @@ class IExecutionProvider {
                                  std::string& dll_path);
 
  private:
+  const std::string type_;
   AllocatorMap allocators_;
 
   // convenience list of the allocators so GetAllocatorList doesn't have to build a new vector each time
diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.h b/onnxruntime/core/providers/cpu/cpu_execution_provider.h
index 43c2fe347b..48fde27bc9 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.h
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.h
@@ -15,16 +15,21 @@ struct CPUExecutionProviderInfo {
 
   explicit CPUExecutionProviderInfo(bool use_arena)
       : create_arena(use_arena) {}
+
   CPUExecutionProviderInfo() = default;
 };
 
-using FuseRuleFn = std::function<void(const onnxruntime::GraphViewer&, std::vector<std::unique_ptr<ComputeCapability>>&)>;
+using FuseRuleFn = std::function<void(const onnxruntime::GraphViewer&,
+                                      std::vector<std::unique_ptr<ComputeCapability>>&)>;
 
 // Logical device representation.
 class CPUExecutionProvider : public IExecutionProvider {
  public:
-  explicit CPUExecutionProvider(const CPUExecutionProviderInfo& info) {
-    DeviceAllocatorRegistrationInfo device_info({OrtMemTypeDefault, [](int) { return std::make_unique<CPUAllocator>(); }, std::numeric_limits<size_t>::max()});
+  explicit CPUExecutionProvider(const CPUExecutionProviderInfo& info)
+      : IExecutionProvider{onnxruntime::kCpuExecutionProvider} {
+    DeviceAllocatorRegistrationInfo device_info{OrtMemTypeDefault,
+                                                [](int) { return std::make_unique<CPUAllocator>(); },
+                                                std::numeric_limits<size_t>::max()};
 #ifdef USE_JEMALLOC
     ORT_UNUSED_PARAMETER(info);
     //JEMalloc already has memory pool, so just use device allocator.
@@ -41,10 +46,6 @@ class CPUExecutionProvider : public IExecutionProvider {
 #endif
   }
 
-  std::string Type() const override {
-    return onnxruntime::kCpuExecutionProvider;
-  }
-
   std::vector<std::unique_ptr<ComputeCapability>> GetCapability(
       const onnxruntime::GraphViewer& graph,
       const std::vector<const KernelRegistry*>& kernel_registries) const override;
@@ -63,7 +64,7 @@ class CPUExecutionProvider : public IExecutionProvider {
 
   void InsertFusedRules(FuseRuleFn rule);
 
- protected:
+ private:
   std::vector<FuseRuleFn> fuse_rules_;
 };
 }  // namespace onnxruntime
diff --git a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
index 51ed64a9f8..c3b7aae053 100644
--- a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
+++ b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
@@ -54,7 +54,7 @@ CUDAExecutionProvider::PerThreadContext::~PerThreadContext() {
 }
 
 CUDAExecutionProvider::CUDAExecutionProvider(const CUDAExecutionProviderInfo& info)
-    : device_id_(info.device_id) {
+    : IExecutionProvider{onnxruntime::kCudaExecutionProvider}, device_id_(info.device_id) {
   CUDA_CALL_THROW(cudaSetDevice(device_id_));
   // create streams, default is nullptr
   streams_[kCudaStreamDefault] = nullptr;
diff --git a/onnxruntime/core/providers/cuda/cuda_execution_provider.h b/onnxruntime/core/providers/cuda/cuda_execution_provider.h
index 60bd5bc9cd..e9f3848e8f 100644
--- a/onnxruntime/core/providers/cuda/cuda_execution_provider.h
+++ b/onnxruntime/core/providers/cuda/cuda_execution_provider.h
@@ -32,10 +32,6 @@ class CUDAExecutionProvider : public IExecutionProvider {
 
   AllocatorPtr GetAllocator(int id, OrtMemType mem_type = OrtMemTypeDefault) const override;
 
-  std::string Type() const override {
-    return onnxruntime::kCudaExecutionProvider;
-  }
-
   Status Sync() const override;
 
   Status OnRunStart() override;
diff --git a/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.cc b/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.cc
index 1a6f1b8844..6163c80266 100644
--- a/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.cc
+++ b/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.cc
@@ -32,7 +32,8 @@ ONNX_OPERATOR_KERNEL_EX(
 
 }  // namespace mkl_dnn
 
-MKLDNNExecutionProvider::MKLDNNExecutionProvider(const MKLDNNExecutionProviderInfo& /*info*/) {
+MKLDNNExecutionProvider::MKLDNNExecutionProvider(const MKLDNNExecutionProviderInfo& /*info*/)
+      : IExecutionProvider{onnxruntime::kMklDnnExecutionProvider} {
   DeviceAllocatorRegistrationInfo default_allocator_info({OrtMemTypeDefault,
                                                           [](int) { return std::make_unique<CPUAllocator>(std::make_unique<OrtAllocatorInfo>(MKLDNN, OrtAllocatorType::OrtDeviceAllocator, 0, OrtMemTypeDefault)); }, std::numeric_limits<size_t>::max()});
   InsertAllocator(CreateAllocator(default_allocator_info));
diff --git a/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.h b/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.h
index a85446ea1b..359f9b96be 100644
--- a/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.h
+++ b/onnxruntime/core/providers/mkldnn/mkldnn_execution_provider.h
@@ -34,10 +34,6 @@ class MKLDNNExecutionProvider : public IExecutionProvider {
   explicit MKLDNNExecutionProvider(const MKLDNNExecutionProviderInfo& info);
   virtual ~MKLDNNExecutionProvider();
 
-  std::string Type() const override {
-    return onnxruntime::kMklDnnExecutionProvider;
-  }
-
   Status CopyTensor(const Tensor& src, Tensor& dst) const override;
 
   const void* GetExecutionHandle() const noexcept override {
diff --git a/onnxruntime/test/framework/dummy_provider.h b/onnxruntime/test/framework/dummy_provider.h
index 818146712a..c434a3a6dd 100644
--- a/onnxruntime/test/framework/dummy_provider.h
+++ b/onnxruntime/test/framework/dummy_provider.h
@@ -14,14 +14,10 @@ class DummyExecutionProvider : public IExecutionProvider {
   static constexpr const char* kDummyExecutionProviderType = "DummyExecutionProvider";
 
  public:
-  DummyExecutionProvider() {
+  DummyExecutionProvider() : IExecutionProvider{kDummyExecutionProviderType} {
     InsertAllocator(std::make_unique<DummyAllocator>());
   }
 
-  std::string Type() const override {
-    return kDummyExecutionProviderType;
-  }
-
   Status CopyTensor(const Tensor& src, Tensor& dst) const override {
     // we can 'copy' from anything we allocated to/from CPU
     ORT_ENFORCE(strcmp(dst.Location().name, DummyAllocator::kDummyAllocator) == 0 ||
diff --git a/onnxruntime/test/framework/inference_session_test.cc b/onnxruntime/test/framework/inference_session_test.cc
index 1a21a2a060..a84e449dc5 100644
--- a/onnxruntime/test/framework/inference_session_test.cc
+++ b/onnxruntime/test/framework/inference_session_test.cc
@@ -77,7 +77,7 @@ std::shared_ptr<KernelRegistry> GetFusedKernelRegistry() {
 
 class FuseExecutionProvider : public IExecutionProvider {
  public:
-  explicit FuseExecutionProvider() {
+  explicit FuseExecutionProvider() : IExecutionProvider{kFuseExecutionProvider} {
     DeviceAllocatorRegistrationInfo device_info({OrtMemTypeDefault,
                                                  [](int) { return std::make_unique<CPUAllocator>(); },
                                                  std::numeric_limits<size_t>::max()});
@@ -120,11 +120,8 @@ class FuseExecutionProvider : public IExecutionProvider {
   const void* GetExecutionHandle() const noexcept override {
     return nullptr;
   }
-
-  std::string Type() const override {
-    return "FuseExecutionProvider";
-  }
 };
+
 namespace test {
 static void VerifyOutputs(const std::vector<MLValue>& fetches,
                           const std::vector<int64_t>& expected_dims,
diff --git a/onnxruntime/test/framework/op_kernel_test.cc b/onnxruntime/test/framework/op_kernel_test.cc
index 032b9a34c6..e2e26e44a6 100644
--- a/onnxruntime/test/framework/op_kernel_test.cc
+++ b/onnxruntime/test/framework/op_kernel_test.cc
@@ -12,18 +12,15 @@
 #include "test_utils.h"
 using namespace std;
 using namespace ONNX_NAMESPACE;
-using namespace ::onnxruntime::logging;
 
 namespace onnxruntime {
+using namespace logging;
+
 namespace test {
 
 class XPUExecutionProvider : public IExecutionProvider {
  public:
-  XPUExecutionProvider() = default;
-
-  std::string Type() const override {
-    return onnxruntime::kCpuExecutionProvider;
-  }
+  XPUExecutionProvider() : IExecutionProvider{onnxruntime::kCpuExecutionProvider} {}
 
   Status CopyTensor(const Tensor& src, Tensor& dst) const override {
     ORT_UNUSED_PARAMETER(src);