From b42cb619044df502a90fcbde1a326948719e0909 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 20 Dec 2019 19:53:51 -0800
Subject: [PATCH 01/75] Packaging pipeline changes for VS 2019 (#2711)

---
 cmake/CMakeLists.txt                          |   7 +-
 ....OnnxRuntime.EndToEndTests.RunCapi.vcxproj |  12 +-
 ...xRuntime.Gpu.EndToEndTests.RunCapi.vcxproj |  10 +-
 tools/ci_build/build.py                       |   3 -
 .../azure-pipelines-py-packaging.yml          |  50 +++--
 .../c-api-packaging-pipelines.yml             |  88 +++++---
 .../nuget/cpu-esrp-pipeline.yml               |   1 -
 .../nuget/cpu-mklml-esrp-pipeline.yml         |   3 +-
 .../nuget/cpu-mklml-pipeline.yml              |   3 +-
 .../cpu-nocontribops-arm64-esrp-pipeline.yml  |   3 +-
 .../nuget/cpu-nocontribops-arm64-pipeline.yml |   3 +-
 .../nuget/gpu-esrp-pipeline.yml               |   2 +-
 .../azure-pipelines/nuget/gpu-pipeline.yml    |   3 +-
 .../nuget/templates/cpu-mklml.yml             |   5 +-
 .../templates/cpu-nocontribops-arm64.yml      |  34 +--
 .../azure-pipelines/nuget/templates/cpu.yml   |  17 +-
 .../azure-pipelines/nuget/templates/gpu.yml   |  24 +-
 .../nuget/templates/test_all_os.yml           |   2 +-
 .../nuget/templates/test_win.yml              |   4 +-
 .../azure-pipelines/post-merge-jobs.yml       |  69 ++++--
 .../azure-pipelines/templates/win-ci-2019.yml |  18 +-
 .../azure-pipelines/templates/win-ci.yml      |   2 +-
 .../templates/win-x86-ci-2019.yml             | 190 ----------------
 .../azure-pipelines/templates/win-x86-ci.yml  | 205 ------------------
 .../templates/windows-code-coverage-steps.yml |  10 +-
 .../github/windows/run_OpenCppCoverage.ps1    |   8 +-
 26 files changed, 225 insertions(+), 551 deletions(-)
 delete mode 100644 tools/ci_build/github/azure-pipelines/templates/win-x86-ci-2019.yml
 delete mode 100644 tools/ci_build/github/azure-pipelines/templates/win-x86-ci.yml

diff --git a/cmake/CMakeLists.txt b/cmake/CMakeLists.txt
index 69fad352b9d27..655b14792f731 100644
--- a/cmake/CMakeLists.txt
+++ b/cmake/CMakeLists.txt
@@ -218,12 +218,7 @@ else()
     string(APPEND CMAKE_C_FLAGS_RELEASE " -march=native -mtune=native")
     string(APPEND CMAKE_CXX_FLAGS_RELWITHDEBINFO " -march=native -mtune=native")
     string(APPEND CMAKE_C_FLAGS_RELWITHDEBINFO " -march=native -mtune=native")
-  endif()
-  if(onnxruntime_BUILD_x86)
-    set (CMAKE_SYSTEM_PROCESSOR "x86")
-    set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -msse2 -mfpmath=sse -Wno-narrowing")
-    set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse2 -mfpmath=sse -Wno-narrowing")
-  endif()
+  endif()  
 endif()
 
 if (${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.EndToEndTests.RunCapi.vcxproj b/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.EndToEndTests.RunCapi.vcxproj
index eb576614c8732..3fb20d386bb2e 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.EndToEndTests.RunCapi.vcxproj
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.EndToEndTests.RunCapi.vcxproj
@@ -1,13 +1,11 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Project DefaultTargets="Build" ToolsVersion="15.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
-  
   <PropertyGroup>
     <OnnxRuntimeCsharpRoot>$(MSBuildThisFileDirectory)..\..</OnnxRuntimeCsharpRoot>
     <PackageName Condition="'$(PackageName)' == ''">Microsoft.ML.OnnxRuntime</PackageName>
     <OnnxRuntimeSampleCode Condition="'$(OnnxRuntimeSampleCode)' == ''">C_Api_Sample.cpp</OnnxRuntimeSampleCode>
   </PropertyGroup>
   <Import Project="..\packages\$(PackageName).$(CurrentOnnxRuntimeVersion)\build\native\$(PackageName).props" Condition="Exists('..\packages\$(PackageName).$(CurrentOnnxRuntimeVersion)\build\native\$(PackageName).props')" />
-
   <ItemGroup Label="ProjectConfigurations">
     <ProjectConfiguration Include="Debug|x64">
       <Configuration>Debug</Configuration>
@@ -17,7 +15,7 @@
       <Configuration>Release</Configuration>
       <Platform>x64</Platform>
     </ProjectConfiguration>
-	<ProjectConfiguration Include="Debug|Win32">
+    <ProjectConfiguration Include="Debug|Win32">
       <Configuration>Debug</Configuration>
       <Platform>x86</Platform>
     </ProjectConfiguration>
@@ -36,13 +34,13 @@
   <PropertyGroup Condition="'$(Configuration)'=='Debug'" Label="Configuration">
     <ConfigurationType>Application</ConfigurationType>
     <UseDebugLibraries>true</UseDebugLibraries>
-    <PlatformToolset>v141</PlatformToolset>
+    <PlatformToolset>v142</PlatformToolset>
     <CharacterSet>Unicode</CharacterSet>
   </PropertyGroup>
   <PropertyGroup Condition="'$(Configuration)'=='Release'" Label="Configuration">
     <ConfigurationType>Application</ConfigurationType>
     <UseDebugLibraries>false</UseDebugLibraries>
-    <PlatformToolset>v141</PlatformToolset>
+    <PlatformToolset>v142</PlatformToolset>
     <WholeProgramOptimization>true</WholeProgramOptimization>
     <CharacterSet>Unicode</CharacterSet>
   </PropertyGroup>
@@ -112,7 +110,7 @@
     <ClCompile Include="$(OnnxRuntimeSampleCode)" />
   </ItemGroup>
   <ItemGroup>
-   <None Include="$(OnnxRuntimeCSharpRoot)\testdata\*">
+    <None Include="$(OnnxRuntimeCSharpRoot)\testdata\*">
       <CopyToOutputDirectory>Always</CopyToOutputDirectory>
       <Visible>false</Visible>
     </None>
@@ -129,4 +127,4 @@
     <Error Condition="!Exists('..\packages\$(PackageName).$(CurrentOnnxRuntimeVersion)\build\native\$(PackageName).props')" Text="$([System.String]::Format('$(ErrorText)', '..\packages\$(PackageName).$(CurrentOnnxRuntimeVersion)\build\native\$(PackageName).props'))" />
     <Error Condition="!Exists('..\packages\$(PackageName).$(CurrentOnnxRuntimeVersion)\build\native\$(PackageName).targets')" Text="$([System.String]::Format('$(ErrorText)', '..\packages\$(PackageName).$(CurrentOnnxRuntimeVersion)\build\native\$(PackageName).targets'))" />
   </Target>
-</Project>
+</Project>
\ No newline at end of file
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.Gpu.EndToEndTests.RunCapi.vcxproj b/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.Gpu.EndToEndTests.RunCapi.vcxproj
index cd93ff64ff455..ea4701e9fedc1 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.Gpu.EndToEndTests.RunCapi.vcxproj
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests.Capi/Microsoft.ML.OnnxRuntime.Gpu.EndToEndTests.RunCapi.vcxproj
@@ -1,11 +1,9 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Project DefaultTargets="Build" ToolsVersion="15.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
-  
   <PropertyGroup>
     <OnnxRuntimeCsharpRoot>$(MSBuildThisFileDirectory)..\..</OnnxRuntimeCsharpRoot>
   </PropertyGroup>
   <Import Project="..\packages\Microsoft.ML.OnnxRuntime.Gpu.$(CurrentOnnxRuntimeVersion)\build\native\Microsoft.ML.OnnxRuntime.Gpu.props" Condition="Exists('..\packages\Microsoft.ML.OnnxRuntime.Gpu.$(CurrentOnnxRuntimeVersion)\build\native\Microsoft.ML.OnnxRuntime.Gpu.props')" />
-
   <ItemGroup Label="ProjectConfigurations">
     <ProjectConfiguration Include="Debug|x64">
       <Configuration>Debug</Configuration>
@@ -26,13 +24,13 @@
   <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
     <ConfigurationType>Application</ConfigurationType>
     <UseDebugLibraries>true</UseDebugLibraries>
-    <PlatformToolset>v141</PlatformToolset>
+    <PlatformToolset>v142</PlatformToolset>
     <CharacterSet>Unicode</CharacterSet>
   </PropertyGroup>
   <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
     <ConfigurationType>Application</ConfigurationType>
     <UseDebugLibraries>false</UseDebugLibraries>
-    <PlatformToolset>v141</PlatformToolset>
+    <PlatformToolset>v142</PlatformToolset>
     <WholeProgramOptimization>true</WholeProgramOptimization>
     <CharacterSet>Unicode</CharacterSet>
   </PropertyGroup>
@@ -90,7 +88,7 @@
     <ClCompile Include="C_Api_Sample.cpp" />
   </ItemGroup>
   <ItemGroup>
-   <None Include="$(OnnxRuntimeCSharpRoot)\testdata\*">
+    <None Include="$(OnnxRuntimeCSharpRoot)\testdata\*">
       <CopyToOutputDirectory>Always</CopyToOutputDirectory>
       <Visible>false</Visible>
     </None>
@@ -107,4 +105,4 @@
     <Error Condition="!Exists('..\packages\Microsoft.ML.OnnxRuntime.Gpu.$(CurrentOnnxRuntimeVersion)\build\native\Microsoft.ML.OnnxRuntime.Gpu.props')" Text="$([System.String]::Format('$(ErrorText)', '..\packages\Microsoft.ML.OnnxRuntime.Gpu.$(CurrentOnnxRuntimeVersion)\build\native\Microsoft.ML.OnnxRuntime.Gpu.props'))" />
     <Error Condition="!Exists('..\packages\Microsoft.ML.OnnxRuntime.Gpu.$(CurrentOnnxRuntimeVersion)\build\native\Microsoft.ML.OnnxRuntime.Gpu.targets')" Text="$([System.String]::Format('$(ErrorText)', '..\packages\Microsoft.ML.OnnxRuntime.Gpu.$(CurrentOnnxRuntimeVersion)\build\native\Microsoft.ML.OnnxRuntime.Gpu.targets'))" />
   </Target>
-</Project>
+</Project>
\ No newline at end of file
diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index 629ced34f51b8..b07d26408d15a 100755
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -141,7 +141,6 @@ def parse_arguments():
     parser.add_argument("--use_tvm", action="store_true", help="Build with tvm")
     parser.add_argument("--use_openmp", action='store_true', help="Build with OpenMP.")
     parser.add_argument("--use_llvm", action="store_true", help="Build tvm with llvm")
-    parser.add_argument("--use_eigenthreadpool", action="store_true", help="Build with eigenthreadpool")
     parser.add_argument("--enable_msinternal", action="store_true", help="Enable for Microsoft internal builds only.")
     parser.add_argument("--llvm_path", help="Path to llvm dir")
     parser.add_argument("--use_brainslice", action="store_true", help="Build with brain slice")
@@ -318,13 +317,11 @@ def generate_build_tree(cmake_path, source_dir, build_dir, cuda_home, cudnn_home
                  "-Donnxruntime_ENABLE_MICROSOFT_INTERNAL=" + ("ON" if args.enable_msinternal else "OFF"),
                  "-Donnxruntime_USE_BRAINSLICE=" + ("ON" if args.use_brainslice else "OFF"),
                  "-Donnxruntime_USE_NUPHAR=" + ("ON" if args.use_nuphar else "OFF"),
-                 "-Donnxruntime_USE_EIGEN_THREADPOOL=" + ("ON" if args.use_eigenthreadpool else "OFF"),
                  "-Donnxruntime_USE_TENSORRT=" + ("ON" if args.use_tensorrt else "OFF"),
                  "-Donnxruntime_TENSORRT_HOME=" + (tensorrt_home if args.use_tensorrt else ""),
                   # By default - we currently support only cross compiling for ARM/ARM64 (no native compilation supported through this script)
                  "-Donnxruntime_CROSS_COMPILING=" + ("ON" if args.arm64 or args.arm else "OFF"),
                  "-Donnxruntime_BUILD_SERVER=" + ("ON" if args.build_server else "OFF"),
-                 "-Donnxruntime_BUILD_x86=" + ("ON" if args.x86 else "OFF"),
                   # nGraph and TensorRT providers currently only supports full_protobuf option.
                  "-Donnxruntime_USE_FULL_PROTOBUF=" + ("ON" if args.use_full_protobuf or args.use_ngraph or args.use_tensorrt or args.build_server or args.gen_doc else "OFF"),
                  "-Donnxruntime_DISABLE_CONTRIB_OPS=" + ("ON" if args.disable_contrib_ops else "OFF"),
diff --git a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
index 0b925ed087878..4054842b70eea 100644
--- a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
+++ b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
@@ -174,13 +174,13 @@ jobs:
     displayName: 'BUILD'
     inputs:
       scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
-      arguments: '--config $(BuildConfig) --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_generator "Visual Studio 16 2019" --build_wheel --use_featurizers --enable_onnx_tests --parallel'
+      arguments: '--config RelWithDebInfo --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_generator "Visual Studio 16 2019" --build_wheel --use_featurizers --enable_onnx_tests --parallel'
       workingDirectory: '$(Build.BinariesDirectory)'     
   
   - task: CopyFiles@2
     displayName: 'Copy Python Wheel to: $(Build.ArtifactStagingDirectory)'
     inputs:
-      SourceFolder: '$(Build.BinariesDirectory)\$(BuildConfig)'
+      SourceFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
       Contents: '**\dist\*.whl'
       TargetFolder: '$(Build.ArtifactStagingDirectory)'
 
@@ -198,11 +198,11 @@ jobs:
 - job:  Windows_py_GPU_Wheels
   workspace:
     clean: all
-  pool: Win-GPU-CUDA10
+  pool: 'Win-GPU-2019'
   timeoutInMinutes:  120  
   variables:
-    buildDirectory: '$(Build.SourcesDirectory)\build'
     CUDA_VERSION: '10.0'
+    EnvSetupScript: setup_env.bat
   strategy:
     matrix:
       Python35:
@@ -217,30 +217,39 @@ jobs:
         versionSpec: $(python.version) 
         addToPath: true 
         architecture: 'x64'
+
     - task: BatchScript@1
-      displayName: 'Setup VS2017 env vars'
+      displayName: 'setup env'
       inputs:
-        filename: 'C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvars64.bat'
-        arguments: 'amd64'
+        filename: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\$(EnvSetupScript)'
         modifyEnvironment: true
-    - task: PowerShell@1
-      displayName: 'Set CUDA path'
+        workingFolder: '$(Build.BinariesDirectory)'
+
+
+    - script: |
+       python -m pip install -q pyopenssl setuptools wheel numpy     
+      workingDirectory: '$(Build.BinariesDirectory)'
+      displayName: 'Install python modules'
+
+    - template: templates/set-test-data-variables-step.yml
+    - task: PythonScript@0
+      displayName: 'Download test data'
       inputs:
-        scriptName: 'tools/ci_build/github/windows/set_cuda_path.ps1'
-        arguments: '-CudaMsbuildPath C:\local\cudaMsbuildIntegration-10.0.130-win10 -CudaVersion 10.0'
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\github\download_test_data.py'
+        arguments: --test_data_url $(TestDataUrl) --build_dir $(Build.BinariesDirectory)
+        workingDirectory: $(Build.BinariesDirectory)
 
-    - task: BatchScript@1
-      displayName: 'Run build script'
+    - task: PythonScript@0
+      displayName: 'build'
       inputs:
-        filename: 'build.bat'
-        arguments: ' --use_cuda --cuda_home="C:\local\cuda_10.0.130_win10"
-      --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda" --build_dir $(buildDirectory) --config Release --build_wheel'
-        workingFolder: "$(Build.SourcesDirectory)"
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
+        arguments: --config RelWithDebInfo --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --cmake_generator "Visual Studio 16 2019" --msvc_toolset 14.16 --build_wheel --enable_onnx_tests --parallel --use_cuda --cuda_version=$(CUDA_VERSION) --cuda_home="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v$(CUDA_VERSION)" --cudnn_home="C:\local\cudnn-$(CUDA_VERSION)-windows10-x64-v7.3.1.20\cuda"
+        workingDirectory: '$(Build.BinariesDirectory)'   
 
     - task: CopyFiles@2
       displayName: 'Copy Python Wheel to: $(Build.ArtifactStagingDirectory)'
       inputs:
-        SourceFolder: '$(buildDirectory)'
+        SourceFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
         Contents: '**\dist\*.whl'
         TargetFolder: '$(Build.ArtifactStagingDirectory)'
 
@@ -249,11 +258,6 @@ jobs:
       inputs:
         ArtifactName: onnxruntime_gpu
 
-    - task: PowerShell@1
-      displayName: 'Clean up CUDA props files'
-      inputs:
-        scriptName: 'tools/ci_build/github/windows/clean_up_cuda_prop_files.ps1'
-        arguments: '-CudaVersion 10.0'
     - task: ms.vss-governance-buildtask.governance-build-task-component-detection.ComponentGovernanceComponentDetection@0
       displayName: 'Component Detection'
 
diff --git a/tools/ci_build/github/azure-pipelines/c-api-packaging-pipelines.yml b/tools/ci_build/github/azure-pipelines/c-api-packaging-pipelines.yml
index 9f18e3d0d3dbd..b04222df0a45e 100644
--- a/tools/ci_build/github/azure-pipelines/c-api-packaging-pipelines.yml
+++ b/tools/ci_build/github/azure-pipelines/c-api-packaging-pipelines.yml
@@ -93,9 +93,6 @@ jobs:
         msbuildArch: amd64
         msbuildPlatform: x64
         buildparameter:
-  variables:
-    buildDirectory: '$(Build.BinariesDirectory)'
-    buildArch: 'x64'
 
   steps:
     - task: UsePythonVersion@0
@@ -161,49 +158,78 @@ jobs:
         commitId: $(OnnxRuntimeGitCommitHash)
     - template: templates/clean-agent-build-directory-step.yml
 
-- job: Windows_Packaging_GPU_x64
+
+- job: Windows_Packaging_GPU
   workspace:
     clean: all
-  timeoutInMinutes:  120    
-  pool: 'Win-GPU-CUDA10'
+  pool: 'Win-GPU-2019'
+  timeoutInMinutes:  120
   variables:
-    buildDirectory: '$(Build.BinariesDirectory)'
-    buildConfig: 'RelWithDebInfo'
-    buildArch: 'x64'
-
+    EnvSetupScript: setup_env.bat
+    buildArch: x64
+    msbuildArch: amd64
+    msbuildPlatform: x64
+    CUDA_VERSION: '10.0'
+    buildparameter: --msvc_toolset 14.16 --use_cuda --cuda_version=$(CUDA_VERSION) --cuda_home="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v$(CUDA_VERSION)" --cudnn_home="C:\local\cudnn-$(CUDA_VERSION)-windows10-x64-v7.3.1.20\cuda"
   steps:
+    - task: UsePythonVersion@0
+      inputs: 
+        versionSpec: '3.7' 
+        addToPath: true 
+        architecture: $(buildArch)
+
+    - task: BatchScript@1
+      displayName: 'setup env'
+      inputs:
+        filename: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\$(EnvSetupScript)'
+        modifyEnvironment: true
+        workingFolder: '$(Build.BinariesDirectory)'
+
+    - script: |
+       python -m pip install -q pyopenssl setuptools wheel numpy     
+      workingDirectory: '$(Build.BinariesDirectory)'
+      displayName: 'Install python modules' 
+    
     - template: templates/set-test-data-variables-step.yml
     - template: templates/set-version-number-variables-step.yml
+    - task: PythonScript@0
+      displayName: 'Download test data'
+      inputs:
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\github\download_test_data.py'
+        arguments: --test_data_url $(TestDataUrl) --build_dir $(Build.BinariesDirectory)
+        workingDirectory: $(Build.BinariesDirectory) 
 
-    - template: templates/windows-build-tools-setup-steps.yml
-      parameters:
-        EnvSetupScript: 'setup_env.bat'
-        buildArch: 'amd64'  # amd64 is needed for vcvars target arch
-        setVcvars: true
 
-    - task: PowerShell@1
-      displayName: 'Set CUDA path'
+    - task: PythonScript@0
+      displayName: 'Generate cmake config'
       inputs:
-        scriptName: 'tools/ci_build/github/windows/set_cuda_path.ps1'
-        arguments: '-CudaMsbuildPath C:\local\cudaMsbuildIntegration-10.0.130-win10 -CudaVersion 10.0'
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
+        arguments: '--config RelWithDebInfo --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --update --cmake_generator "Visual Studio 16 2019" --use_openmp --build_shared_lib --enable_onnx_tests $(buildparameter)'
+        workingDirectory: '$(Build.BinariesDirectory)'
 
-    - task: CmdLine@2
-      displayName: 'Build and Test OnnxRuntime'
+ 
+    - task: VSBuild@1
+      displayName: 'Build'
       inputs:
-        script: |
-          $(Build.BinariesDirectory)\packages\python\python.exe $(Build.SourcesDirectory)\tools\ci_build\build.py --config $(buildConfig) --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --enable_onnx_tests --use_cuda --cuda_version=10.0 --cuda_home="C:\local\cuda_10.0.130_win10" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"
+        solution: '$(Build.BinariesDirectory)\RelWithDebInfo\onnxruntime.sln'
+        platform: $(msbuildPlatform)
+        configuration: RelWithDebInfo
+        msbuildArchitecture: $(buildArch)
+        maximumCpuCount: true
+        logProjectEvents: true
+        workingFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
+        createLogFile: true
+
+    - task: PythonScript@0
+      displayName: 'test'
+      inputs:
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
+        arguments: '--config RelWithDebInfo --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --test --cmake_generator "Visual Studio 16 2019" --use_openmp --build_shared_lib --enable_onnx_tests $(buildparameter)'
         workingDirectory: '$(Build.BinariesDirectory)'
 
     - template: templates/c-api-artifacts-package-and-publish-steps-windows.yml
       parameters:
-        buildConfig: $(buildConfig)
+        buildConfig: RelWithDebInfo
         artifactName: 'onnxruntime-win-$(buildArch)-gpu-$(OnnxRuntimeVersion)'
         commitId: $(OnnxRuntimeGitCommitHash)
-
-    - task: PowerShell@1
-      displayName: 'Clean up CUDA props files'
-      inputs:
-        scriptName: 'tools/ci_build/github/windows/clean_up_cuda_prop_files.ps1'
-        arguments: '-CudaVersion 10.0'
-
     - template: templates/clean-agent-build-directory-step.yml
\ No newline at end of file
diff --git a/tools/ci_build/github/azure-pipelines/nuget/cpu-esrp-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/cpu-esrp-pipeline.yml
index 062ff28790fba..64f82c148364c 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/cpu-esrp-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/cpu-esrp-pipeline.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
diff --git a/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-esrp-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-esrp-pipeline.yml
index aac065f365019..e7d942c4ab57a 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-esrp-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-esrp-pipeline.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -18,5 +17,5 @@ variables:
 jobs: 
 - template: templates/cpu-mklml.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
     DoEsrp: 'true'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-pipeline.yml
index 135469669c7ed..29581701c71e2 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/cpu-mklml-pipeline.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -10,5 +9,5 @@ variables:
 jobs: 
 - template: templates/cpu-mklml.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
     DoEsrp: 'false'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-esrp-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-esrp-pipeline.yml
index 56c9a86167b74..773295e3d23c3 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-esrp-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-esrp-pipeline.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -18,5 +17,5 @@ variables:
 jobs:
 - template: templates/cpu-nocontribops-arm64.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
     DoEsrp: 'true'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-pipeline.yml
index b09b21ef1ca44..ba0a94e3bd105 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/cpu-nocontribops-arm64-pipeline.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -10,5 +9,5 @@ variables:
 jobs:
 - template: templates/cpu-nocontribops-arm64.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
     DoEsrp: 'false'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/gpu-esrp-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/gpu-esrp-pipeline.yml
index 82b795c05fcc4..917c83a6393b1 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/gpu-esrp-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/gpu-esrp-pipeline.yml
@@ -19,5 +19,5 @@ variables:
 jobs: 
 - template: templates/gpu.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-GPU-2019'
     DoEsrp: 'true'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/gpu-pipeline.yml b/tools/ci_build/github/azure-pipelines/nuget/gpu-pipeline.yml
index ef040845eefe1..5057490121b9a 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/gpu-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/gpu-pipeline.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU-CUDA10'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -11,5 +10,5 @@ variables:
 jobs: 
 - template: templates/gpu.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-GPU-2019'
     DoEsrp: 'false'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml
index 5d25a5d29f3e8..afee0189d90d4 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -13,6 +12,10 @@ jobs:
     AgentPool : 'Win-CPU-2019'
     JobName: 'Windows_CI_Dev'
     BuildCommand:  '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_mklml --build_shared_lib --enable_onnx_tests  --cmake_generator "Visual Studio 16 2019"'
+    BuildArch: 'x64'
+    msbuildArchitecture: 'amd64'
+    EnvSetupScript: 'setup_env.bat'
+    sln_platform: 'x64'
     DoDebugBuild: 'false'
     DoNugetPack : 'true'
     DoCompliance: 'false'
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml
index 6aa762839af33..b394feccbb997 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml
@@ -1,21 +1,24 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
 parameters:
   DoEsrp: 'false'
 
-jobs:
-- template: ../../templates/win-ci.yml
+- template: ../../templates/win-ci-2019.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
+    ArtifactName: 'drop-nuget'
     JobName: 'Windows_CI_Dev'
-    BuildCommand:  '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_path $(Build.BinariesDirectory)\cmake\bin\cmake.exe --ctest_path $(Build.BinariesDirectory)\cmake\bin\ctest.exe --disable_contrib_ops --enable_msvc_static_runtime --build_shared_lib --build_csharp --enable_onnx_tests'
+    BuildCommand:  '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019" --disable_contrib_ops --enable_msvc_static_runtime'
+    BuildArch: 'x64'
+    msbuildArchitecture: 'amd64'
+    EnvSetupScript: 'setup_env.bat'
+    sln_platform: 'x64'
     DoDebugBuild: 'false'
     DoNugetPack : 'true'
-    DoCompliance: 'false'
+    DoCompliance: ${{ parameters.DoCompliance }}
     DoEsrp: ${{ parameters.DoEsrp }}
     NuPackScript: |
      msbuild $(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.proj /p:Configuration=RelWithDebInfo /t:CreatePackage
@@ -23,11 +26,16 @@ jobs:
      mkdir $(Build.ArtifactStagingDirectory)\testdata
      copy $(Build.BinariesDirectory)\RelWithDebInfo\RelWithDebInfo\custom_op_library.* $(Build.ArtifactStagingDirectory)\testdata
 
-- template: ../../templates/win-x86-ci.yml
+- template: ../../templates/win-ci-2019.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
+    ArtifactName: 'drop-win-x86-zip'
     JobName: 'Windows_CI_Dev_x86'
-    BuildCommand: '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_path $(Build.BinariesDirectory)\cmake\bin\cmake.exe --ctest_path $(Build.BinariesDirectory)\cmake\bin\ctest.exe --disable_contrib_ops --enable_msvc_static_runtime --build_shared_lib  --build_csharp --enable_onnx_tests --x86'
+    BuildCommand: '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --enable_onnx_tests --x86 --use_telemetry --cmake_generator "Visual Studio 16 2019" --disable_contrib_ops --enable_msvc_static_runtime'
+    BuildArch: 'x86'
+    msbuildArchitecture: 'x86'
+    EnvSetupScript: 'setup_env_x86.bat'
+    sln_platform: 'Win32'
     DoDebugBuild: 'false'
     DoNugetPack : 'true'
     DoCompliance: 'false'
@@ -42,7 +50,7 @@ jobs:
 
 - template: ../../templates/win-ci-arm.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
     JobName: 'Windows_Arm64_Dev'
     BuildCommand:  '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_path $(Build.BinariesDirectory)\cmake\bin\cmake.exe --ctest_path $(Build.BinariesDirectory)\cmake\bin\ctest.exe --disable_contrib_ops --enable_msvc_static_runtime --build_shared_lib --arm64'
     DoDebugBuild: 'false'
@@ -76,7 +84,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-centos6 /bin/bash -c "/usr/bin/python3.6 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --enable_onnx_tests --disable_contrib_ops && cd /build/Release && make install DESTDIR=/build/linux-x64"
+          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-centos6 /bin/bash -c "/usr/bin/python3.6 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --enable_onnx_tests --disable_contrib_ops && cd /build/Release && make install DESTDIR=/build/linux-x64"
         workingDirectory: $(Build.SourcesDirectory)
     - script: |
        set -e -x
@@ -102,7 +110,7 @@ jobs:
   parameters:
     AgentPool : $(AgentPoolMacOS)
     JobName: 'MacOS_CI_Dev'
-    BuildCommand: 'python3 $(Build.SourcesDirectory)/tools/ci_build/build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --parallel --build_shared_lib --disable_contrib_ops --use_openmp --enable_onnx_tests --config RelWithDebInfo'
+    BuildCommand: 'python3 $(Build.SourcesDirectory)/tools/ci_build/build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --parallel --build_shared_lib --disable_contrib_ops --enable_onnx_tests --config RelWithDebInfo'
     DoNugetPack : 'true'
     NuPackScript: |
      set -e -x
@@ -123,7 +131,7 @@ jobs:
 - job: NuGet_Packaging
   workspace:
     clean: all
-  pool: $(AgentPoolWin)
+  pool: 'Win-CPU-2019'
   dependsOn:
   - Windows_CI_Dev
   - Windows_CI_Dev_x86
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
index 2cf05fc127315..95672d0be00a0 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
@@ -5,14 +5,20 @@
 
 parameters:
   DoEsrp: 'false'
+  PackageName: 'Microsoft.ML.OnnxRuntime'
   DoCompliance: 'false'
 
 jobs: 
 - template: ../../templates/win-ci-2019.yml
   parameters:
     AgentPool : 'Win-CPU-2019'
+    ArtifactName: 'drop-nuget'
     JobName: 'Windows_CI_Dev'
     BuildCommand:  '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_openmp --build_shared_lib --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019"'
+    BuildArch: 'x64'
+    msbuildArchitecture: 'amd64'
+    EnvSetupScript: 'setup_env.bat'
+    sln_platform: 'x64'
     DoDebugBuild: 'false'
     DoNugetPack : 'true'
     DoCompliance: ${{ parameters.DoCompliance }}
@@ -23,11 +29,16 @@ jobs:
      mkdir $(Build.ArtifactStagingDirectory)\testdata
      copy $(Build.BinariesDirectory)\RelWithDebInfo\RelWithDebInfo\custom_op_library.* $(Build.ArtifactStagingDirectory)\testdata
 
-- template: ../../templates/win-x86-ci-2019.yml
+- template: ../../templates/win-ci-2019.yml
   parameters:
     AgentPool : 'Win-CPU-2019'
+    ArtifactName: 'drop-win-x86-zip'
     JobName: 'Windows_CI_Dev_x86'
     BuildCommand: '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_openmp --build_shared_lib --enable_onnx_tests --x86 --use_telemetry --cmake_generator "Visual Studio 16 2019"'
+    BuildArch: 'x86'
+    msbuildArchitecture: 'x86'
+    EnvSetupScript: 'setup_env_x86.bat'
+    sln_platform: 'Win32'
     DoDebugBuild: 'false'
     DoNugetPack : 'true'
     DoCompliance: 'false'
@@ -116,7 +127,6 @@ jobs:
     inputs:
       artifactName: 'drop-nuget'
       targetPath: '$(Build.BinariesDirectory)/nuget-artifact'
-    continueOnError: true
 
 
   - task: DownloadPipelineArtifact@0
@@ -124,21 +134,18 @@ jobs:
     inputs:
       artifactName: 'drop-win-x86-zip'
       targetPath: '$(Build.BinariesDirectory)/nuget-artifact'
-    continueOnError: true
 
   - task: DownloadPipelineArtifact@0
     displayName: 'Download Pipeline Artifact - Linux'
     inputs:
       artifactName: 'drop-linux'
       targetPath: '$(Build.BinariesDirectory)/nuget-artifact'
-    continueOnError: true
 
   - task: DownloadPipelineArtifact@0
     displayName: 'Download Pipeline Artifact - MacOS'
     inputs:
       artifactName: 'drop-osx'
       targetPath: '$(Build.BinariesDirectory)/nuget-artifact'
-    continueOnError: true
 
   - template: bundle_dlls.yml
 
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
index b7ae98e38c43d..d34ac37afeb04 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
@@ -1,6 +1,5 @@
 # Defined as pipeline variables
 # variables:
-#   AgentPoolWin : 'Win-CPU-CUDA10'
 #   AgentPoolLinux : 'Linux-CPU'
 #   AgentPoolMacOS : 'macOS-10.13'
 
@@ -9,19 +8,20 @@ parameters:
   PackageName: 'Microsoft.ML.OnnxRuntime.Gpu'
 
 jobs: 
-- template: ../../templates/win-ci.yml
+- template: ../../templates/win-ci-2019.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-GPU-2019'
+    ArtifactName: 'drop-nuget'    
     JobName: 'Windows_CI_GPU_Dev'
-    BuildCommand: '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_path $(Build.BinariesDirectory)\cmake\bin\cmake.exe --ctest_path $(Build.BinariesDirectory)\cmake\bin\ctest.exe  --enable_pybind --build_shared_lib  --build_csharp --enable_onnx_tests --use_cuda --cuda_home="C:\local\cuda_10.0.130_win10" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda" --cuda_version 10.0'
+    BuildCommand: --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019" --msvc_toolset 14.16 --use_cuda --cuda_version=10.0 --cuda_home="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"
+    BuildArch: 'x64'
+    msbuildArchitecture: 'amd64'
+    EnvSetupScript: 'setup_env.bat'    
+    sln_platform: 'x64'
     DoDebugBuild: 'false'
     DoNugetPack : 'true'
     DoCompliance: 'false'
-    DoEsrp: ${{ parameters.DoEsrp }}
-    BuildArch: 'amd64'
-    SetVcvars: 'true'
-    MsbuildArguments: '/m /p:CudaToolkitDir=C:\local\cuda_10.0.130_win10\'
-    EnvSetupScript: 'setup_env_cuda.bat'
+    DoEsrp: ${{ parameters.DoEsrp }}    
     CudaVersion: '10.0'
     OrtPackageId: 'Microsoft.ML.OnnxRuntime.Gpu'
     NuPackScript: |
@@ -65,7 +65,7 @@ jobs:
     - template: ../../templates/clean-agent-build-directory-step.yml
 
 - job: NuGet_Packaging
-  pool: $(AgentPoolWin)
+  pool: 'Win-GPU-2019'
   dependsOn:
   - Windows_CI_GPU_Dev
   - Linux_CI_GPU_Dev
@@ -76,14 +76,12 @@ jobs:
     inputs:
       artifactName: 'drop-nuget'
       targetPath: '$(Build.BinariesDirectory)/nuget-artifact'
-    continueOnError: true
 
   - task: DownloadPipelineArtifact@0
     displayName: 'Download Pipeline Artifact - Linux'
     inputs:
       artifactName: 'drop-linux'
       targetPath: '$(Build.BinariesDirectory)/nuget-artifact'
-    continueOnError: true
 
   - script: |
      pushd $(Build.BinariesDirectory)\nuget-artifact
@@ -122,7 +120,7 @@ jobs:
 
 - template: test_win.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-GPU-2019'
 - template: test_linux.yml
   parameters:
     AgentPool : $(AgentPoolLinux)
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/test_all_os.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/test_all_os.yml
index 8a60f4a2675ef..1eac74b5fa532 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/test_all_os.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/test_all_os.yml
@@ -2,7 +2,7 @@
 jobs:
 - template: test_win.yml
   parameters:
-    AgentPool : $(AgentPoolWin)
+    AgentPool : 'Win-CPU-2019'
 - template: test_linux.yml
   parameters:
     AgentPool : $(AgentPoolLinux)
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/test_win.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/test_win.yml
index e4fc8b2e2d018..c2b426b609528 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/test_win.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/test_win.yml
@@ -25,9 +25,9 @@ jobs:
   - template: win-download-test-data.yml
 
   - task: BatchScript@1
-    displayName: 'Setup VS2017 env vars'
+    displayName: 'Setup VS2019 env vars'
     inputs:
-      filename: 'C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvarsall.bat'
+      filename: 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvarsall.bat'
       arguments: 'amd64'
       modifyEnvironment: true
   
diff --git a/tools/ci_build/github/azure-pipelines/post-merge-jobs.yml b/tools/ci_build/github/azure-pipelines/post-merge-jobs.yml
index 82cb2edab499e..cf3b4506c1f18 100644
--- a/tools/ci_build/github/azure-pipelines/post-merge-jobs.yml
+++ b/tools/ci_build/github/azure-pipelines/post-merge-jobs.yml
@@ -1,35 +1,72 @@
 # Jobs to run on master after every successful merge is committed
 jobs:
 - job: Windows_Debug_CPU_x64
-  pool: 'Win-CPU'
+  pool: 'Win-CPU-2019'
   timeoutInMinutes: 120
   variables:
     - group: dashboard_mysql_secret
     - name: buildDirectory
       value: '$(Build.BinariesDirectory)'
-    - name: buildConfig
+    - name: BuildConfig
       value: 'Debug'
     - name: buildArch
       value: 'x64'
+    - name: EnvSetupScript
+      value: 'setup_env.bat'
 
   steps:
     - template: templates/set-test-data-variables-step.yml
     - template: templates/set-version-number-variables-step.yml
 
-    - template: templates/windows-build-tools-setup-steps.yml
-      parameters:
-        EnvSetupScript: 'setup_env.bat'
-        buildArch: 'amd64'  # amd64 is needed for vcvars target arch 
-        setVcvars: false
+    - task: UsePythonVersion@0
+      inputs: 
+        versionSpec: '3.7' 
+        addToPath: true 
+        architecture: $(buildArch)
 
-    - template: templates/windows-build-and-test-steps.yml
-      parameters:
-        buildAdditionalParams: ' --use_openmp '
-        # --use_openmp is added, similar to the build config of the C-api package distribution.
-        # MLAS(default) with OpenMP is known to produce better performance (latency) than without OpenMP. 
-        buildArch: $(buildArch)
-        msbuildPlatform: $(buildArch)
-        buildConfig: $(buildConfig)
+    - task: BatchScript@1
+      displayName: 'setup env'
+      inputs:
+        filename: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\$(EnvSetupScript)'
+        modifyEnvironment: true
+        workingFolder: '$(Build.BinariesDirectory)'
+
+    - task: PythonScript@0
+      displayName: 'Download test data'
+      inputs:
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\github\download_test_data.py'
+        arguments: --test_data_url $(TestDataUrl) --build_dir $(Build.BinariesDirectory)
+        workingDirectory: $(Build.BinariesDirectory)
+
+
+    - powershell: 'Start-Process -FilePath "$(Build.BinariesDirectory)\installer\opencppcoverage\installer.exe" -ArgumentList ''/SP-'', ''/VERYSILENT'', ''/SUPPRESSMSGBOXES'', ''/NORESTART'', ''/DIR="$(Build.BinariesDirectory)\OpenCppCoverage'' -NoNewWindow -Wait'
+      displayName: 'Run OpenCPPCoverage installer'
+
+    - task: PythonScript@0
+      displayName: 'Generate cmake config'
+      inputs:
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
+        arguments: '--config $(BuildConfig) --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --update --cmake_generator "Visual Studio 16 2019" --use_openmp --build_shared_lib --enable_onnx_tests'
+        workingDirectory: '$(Build.BinariesDirectory)'
+
+    - task: VSBuild@1
+      displayName: 'Build'
+      inputs:
+        solution: '$(Build.BinariesDirectory)\$(BuildConfig)\onnxruntime.sln'
+        platform: 'x64'
+        configuration: $(BuildConfig)
+        msbuildArchitecture: $(buildArch)
+        maximumCpuCount: true
+        logProjectEvents: false
+        workingFolder: '$(Build.BinariesDirectory)\$(BuildConfig)'
+        createLogFile: true    
+
+    - task: PythonScript@0
+      displayName: 'test'
+      inputs:
+        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
+        arguments: '--config $(BuildConfig) --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --test --cmake_generator "Visual Studio 16 2019" --use_openmp --build_shared_lib --enable_onnx_tests'
+        workingDirectory: '$(Build.BinariesDirectory)'
 
     - template: templates/windows-code-coverage-steps.yml
       parameters:
@@ -38,5 +75,3 @@ jobs:
         PublishReport: true
         PostToDashboard: true
 
-    - template: templates/clean-agent-build-directory-step.yml
-
diff --git a/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml b/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml
index c018d6f02089e..d4ae4da1b7ec1 100644
--- a/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml
@@ -8,12 +8,12 @@ parameters:
   ArtifactName: 'drop-nuget'
   DoEsrp: 'false'
   DoTestCoverage: 'false'
-  BuildArch: 'x64'
-  SetVcvars: 'false'
-  msbuildArchitecture: 'amd64'
+  BuildArch: 'x64' # Optional. Options: x86, x64
+  msbuildArchitecture: 'amd64' # Options: x86, amd64
+  sln_platform: 'x64' # Options: Win32, x64
   EnvSetupScript: 'setup_env.bat'
   CudaVersion: ''
-  AgentPool: 'Win-CPU'
+  AgentPool: 'Win-CPU-2019'
   AgentDemands: []
   OrtPackageId: Microsoft.ML.OnnxRuntime
 jobs:
@@ -30,6 +30,9 @@ jobs:
     OnnxRuntimeBuildDirectory: '$(Build.BinariesDirectory)'
     DotNetExe: 'dotnet.exe'
     CUDA_VERSION: ${{ parameters.CudaVersion }}
+    DOTNET_SKIP_FIRST_TIME_EXPERIENCE: true
+    PlatformTarget: ${{ parameters.BuildArch }}
+    RuntimeIdentifier: win-${{ parameters.BuildArch }}
 
   steps:
     - powershell: |
@@ -38,9 +41,12 @@ jobs:
           $length = $env:TELEMETRYGUID.length
           $fileContent = "#define ENABLE_TELEMETRY`n#define TraceLoggingOptionMicrosoftTelemetry() \
             TraceLoggingOptionGroup("+$env:TELEMETRYGUID.substring(1, $length-2)+")"
-          New-Item -Path "$(Build.SourcesDirectory)\include\onnxruntime\core\platform\windows\TraceLoggingConfigPrivate.h" -ItemType "file" -Value "$fileContent" -Force       
+          New-Item -Path "$(Build.SourcesDirectory)\include\onnxruntime\core\platform\windows\TraceLoggingConfigPrivate.h" -ItemType "file" -Value "$fileContent" -Force
+          Write-Output "Enabling TELEMETRY"
         }
       displayName: 'Create TraceLoggingConfigPrivate.h For WinML Telemetry'
+      env:
+        TELEMETRYGUID: $(TELEMETRYGUID)
     - task: UsePythonVersion@0
       inputs: 
         versionSpec: '3.7' 
@@ -71,7 +77,7 @@ jobs:
       displayName: 'Build'
       inputs:
         solution: '$(Build.BinariesDirectory)\RelWithDebInfo\onnxruntime.sln'
-        platform: 'x64'
+        platform: ${{ parameters.sln_platform }}
         configuration: RelWithDebInfo
         msbuildArchitecture: ${{ parameters.BuildArch }}
         maximumCpuCount: true
diff --git a/tools/ci_build/github/azure-pipelines/templates/win-ci.yml b/tools/ci_build/github/azure-pipelines/templates/win-ci.yml
index 92ebe320a26de..033923dc34ebb 100644
--- a/tools/ci_build/github/azure-pipelines/templates/win-ci.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/win-ci.yml
@@ -14,7 +14,7 @@ parameters:
   MsbuildArguments: '/m'
   EnvSetupScript: 'setup_env.bat'
   CudaVersion: ''
-  AgentPool: 'Win-CPU'
+  AgentPool: 'Win-CPU-2019'
   AgentDemands: []
   OrtPackageId: Microsoft.ML.OnnxRuntime
 jobs:
diff --git a/tools/ci_build/github/azure-pipelines/templates/win-x86-ci-2019.yml b/tools/ci_build/github/azure-pipelines/templates/win-x86-ci-2019.yml
deleted file mode 100644
index 167d1c0b2f5de..0000000000000
--- a/tools/ci_build/github/azure-pipelines/templates/win-x86-ci-2019.yml
+++ /dev/null
@@ -1,190 +0,0 @@
-parameters:
-  DoDebugBuild: 'true'
-  DoCompliance: 'false'
-  BuildCommand: ''
-  JobName: 'Windows_CI_Dev_x86'
-  DoNugetPack:  'false'
-  NuPackScript : ''
-  ArtifactName: 'drop-win-x86-zip'
-  DoEsrp: 'false'
-  DoTestCoverage: 'false'
-  BuildArch: 'x86'
-  SetVcvars: 'false'
-  msbuildArchitecture: 'x86'
-  EnvSetupScript: 'setup_env_x86.bat'
-  CudaVersion: ''
-  AgentPool: 'Win-CPU'
-  AgentDemands: []
-  OrtPackageId: Microsoft.ML.OnnxRuntime
-jobs:
-- job: ${{ parameters.JobName }}
-  timeoutInMinutes: 120
-  workspace:
-    clean: all
-  pool:
-    name: ${{ parameters.AgentPool }}
-    demands: ${{ parameters.AgentDemands }}
-  variables:
-    buildDirectory: '$(Build.BinariesDirectory)'
-    BuildCommand: ${{ parameters.BuildCommand }}
-    OnnxRuntimeBuildDirectory: '$(Build.BinariesDirectory)'
-    DotNetExe: 'dotnet.exe'
-    CUDA_VERSION: ${{ parameters.CudaVersion }}
-    PlatformTarget: x86
-    RuntimeIdentifier: win-x86
-
-  steps:
-    - powershell: |
-        if($env:TELEMETRYGUID)
-        {
-          $length = $env:TELEMETRYGUID.length
-          $fileContent = "#define ENABLE_TELEMETRY`n#define TraceLoggingOptionMicrosoftTelemetry() \
-            TraceLoggingOptionGroup("+$env:TELEMETRYGUID.substring(1, $length-2)+")"
-          New-Item -Path "$(Build.SourcesDirectory)\include\onnxruntime\core\platform\windows\TraceLoggingConfigPrivate.h" -ItemType "file" -Value "$fileContent" -Force       
-        }
-      displayName: 'Create TraceLoggingConfigPrivate.h For WinML Telemetry'
-    - task: UsePythonVersion@0
-      inputs: 
-        versionSpec: '3.7' 
-        addToPath: true 
-        architecture: ${{ parameters.BuildArch }}
-
-    - task: BatchScript@1
-      displayName: 'setup env'
-      inputs:
-        filename: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\${{ parameters.EnvSetupScript }}'
-        modifyEnvironment: true
-        workingFolder: '$(Build.BinariesDirectory)'
-
-    - script: |
-       python -m pip install -q pyopenssl setuptools wheel numpy     
-      workingDirectory: '$(Build.BinariesDirectory)'
-      displayName: 'Install python modules'
-
-
-    - task: PythonScript@0
-      displayName: 'Generate cmake config'
-      inputs:
-        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
-        arguments: '$(BuildCommand) --update --config RelWithDebInfo'
-        workingDirectory: '$(Build.BinariesDirectory)'
-
-    - task: VSBuild@1
-      displayName: 'Build'
-      inputs:
-        solution: '$(Build.BinariesDirectory)\RelWithDebInfo\onnxruntime.sln'
-        platform: 'Win32'
-        configuration: RelWithDebInfo
-        msbuildArchitecture: ${{ parameters.BuildArch }}
-        maximumCpuCount: true
-        logProjectEvents: true
-        workingFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
-        createLogFile: true
-        clean: true
-
-    # Build RelWithDebInfo -- this variable required to build C#
-    - script: |
-       @echo ##vso[task.setvariable variable=Configuration]RelWithDebInfo
-
-    - template: set-test-data-variables-step.yml
-
-    - task: NuGetToolInstaller@0
-      displayName: Use Nuget 4.9
-      inputs:
-        versionSpec: 4.9.4   
-
-    - task: PythonScript@0
-      displayName: 'Download test data'
-      inputs:
-        scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\github\download_test_data.py'
-        arguments: --test_data_url $(TestDataUrl) --build_dir $(Build.BinariesDirectory)
-        workingDirectory: $(Build.BinariesDirectory)
-
-    - task: DotNetCoreCLI@2
-      displayName: 'Restore nuget packages'
-      inputs:
-        command: restore
-        projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-        configuration: 'RelWithDebInfo'
-        arguments: '--configuration RelWithDebInfo -p:Platform="Any CPU" -p:OrtPackageId=${{ parameters.OrtPackageId }}'
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'      
-
-    - task: DotNetCoreCLI@2
-      displayName: 'Build C#'    
-      inputs:
-        command: build
-        projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-        configuration: 'RelWithDebInfo'          
-        arguments: '--configuration RelWithDebInfo -p:Platform="Any CPU" -p:OnnxRuntimeBuildDirectory="$(Build.BinariesDirectory)" -p:OrtPackageId=${{ parameters.OrtPackageId }}'
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-
-    - task: DotNetCoreCLI@2
-      displayName: 'Test C#'  
-      inputs:
-        command: test
-        projects: '$(Build.SourcesDirectory)\csharp\test\Microsoft.ML.OnnxRuntime.Tests\Microsoft.ML.OnnxRuntime.Tests.csproj'
-        configuration: 'RelWithDebInfo'          
-        arguments: '--configuration RelWithDebInfo -p:Platform="Any CPU" -p:OnnxRuntimeBuildDirectory="$(Build.BinariesDirectory)" -p:OrtPackageId=${{ parameters.OrtPackageId }}'
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-
-    - script: |
-       mklink  /D /J $(Build.BinariesDirectory)\RelWithDebInfo\models $(Build.BinariesDirectory)\models  
-       DIR dist\ /S /B > wheel_filename_file
-       set /p WHEEL_FILENAME=<wheel_filename_file
-       del wheel_filename_file
-       python.exe -m pip install -q --upgrade %WHEEL_FILENAME%
-       set PATH=%PATH%;$(Build.BinariesDirectory)\RelWithDebInfo\RelWithDebInfo
-       @echo %PATH%
-       python $(Build.SourcesDirectory)\tools\ci_build\build.py $(BuildCommand) --test --config RelWithDebInfo    
-      workingDirectory: '$(Build.BinariesDirectory)\RelWithDebInfo\RelWithDebInfo'
-      displayName: 'Run tests' 
-
-    - task: PublishTestResults@2
-      displayName: 'Publish unit test results'
-      inputs:
-        testResultsFiles: '**\*.results.xml'
-        searchFolder: '$(Build.BinariesDirectory)'
-        testRunTitle: 'Unit Test Run'
-      condition: succeededOrFailed()
-
-    # Nuget packaging if needed
-    - ${{ if eq(parameters['DoNugetPack'], 'true') }}:
-      - task: BatchScript@1
-        displayName: 'Setup VS2017 env vars'
-        inputs:
-          filename: 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvarsall.bat'
-          arguments: ${{ parameters.msbuildArchitecture }}
-          modifyEnvironment: true    
-      # Esrp signing
-      - template: esrp_dll.yml
-        parameters:
-          FolderPath: '$(Build.BinariesDirectory)\RelWithDebInfo'
-          DisplayName: 'ESRP - Sign Native dlls'
-          DoEsrp: ${{ parameters.DoEsrp }}
-
-      - template: esrp_dll.yml
-        parameters:
-          FolderPath: '$(Build.SourcesDirectory)\csharp\src\Microsoft.ML.OnnxRuntime\bin\RelWithDebInfo'
-          DisplayName: 'ESRP - Sign C# dlls'
-          DoEsrp: ${{ parameters.DoEsrp }}
-
-      - script: |
-         ${{ parameters.NuPackScript }}
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-        displayName: 'Create NuGet Package'
-
-      - task: PublishPipelineArtifact@0
-        displayName: 'Publish Pipeline Artifact: drop-nuget'
-        inputs:
-          artifactName: ${{ parameters.ArtifactName }}
-          targetPath: '$(Build.ArtifactStagingDirectory)'
-
-    # Compliance tasks require logs from Debug Build
-    - ${{ if eq(parameters['DoCompliance'], 'true') }}:
-      - template: compliance.yml
-
-    - task: ms.vss-governance-buildtask.governance-build-task-component-detection.ComponentGovernanceComponentDetection@0
-      displayName: 'Component Detection'
-      condition: succeeded()
-
-    - template: clean-agent-build-directory-step.yml
diff --git a/tools/ci_build/github/azure-pipelines/templates/win-x86-ci.yml b/tools/ci_build/github/azure-pipelines/templates/win-x86-ci.yml
deleted file mode 100644
index a58ec9a8266e9..0000000000000
--- a/tools/ci_build/github/azure-pipelines/templates/win-x86-ci.yml
+++ /dev/null
@@ -1,205 +0,0 @@
-parameters:
-  DoDebugBuild: 'true'
-  DoCompliance: 'false'
-  BuildCommand: ''
-  JobName: 'Windows_CI_Dev_x86'
-  DoNugetPack:  'false'
-  NuPackScript : ''
-  ArtifactName: 'drop-win-x86-zip'
-  DoEsrp: 'false'
-  DoTestCoverage: 'false'
-  BuildArch: 'x86'
-  SetVcvars: 'false'
-  msbuildArchitecture: 'x86'
-  MsbuildArguments: '/m'
-  EnvSetupScript: 'setup_env_x86.bat'
-  CudaVersion: ''
-  AgentPool: 'Win-CPU'
-  AgentDemands: []
-
-jobs:
-- job: ${{ parameters.JobName }}
-  workspace:
-    clean: all
-  timeoutInMinutes: 120
-  pool:
-    name: ${{ parameters.AgentPool }}
-    demands: ${{ parameters.AgentDemands }}
-  variables:
-    buildDirectory: '$(Build.BinariesDirectory)'
-    BuildCommand: ${{ parameters.BuildCommand }}
-    OnnxRuntimeBuildDirectory: '$(Build.BinariesDirectory)'
-    PlatformTarget: x86
-    RuntimeIdentifier: win-x86
-
-  steps:
-    - powershell: |
-        if($env:TELEMETRYGUID)
-        {
-          $length = $env:TELEMETRYGUID.length
-          $fileContent = "#define ENABLE_TELEMETRY`n#define TraceLoggingOptionMicrosoftTelemetry() \
-            TraceLoggingOptionGroup("+$env:TELEMETRYGUID.substring(1, $length-2)+")"
-          New-Item -Path "$(Build.SourcesDirectory)\include\onnxruntime\core\platform\windows\TraceLoggingConfigPrivate.h" -ItemType "file" -Value "$fileContent" -Force
-        }
-      displayName: 'Create TraceLoggingConfigPrivate.h For WinML Telemetry'
-    - template: set-test-data-variables-step.yml
-    - template: windows-build-tools-setup-steps.yml
-      parameters:
-        EnvSetupScript: ${{ parameters.EnvSetupScript }}
-        buildArch: ${{ parameters.msbuildArchitecture }}
-        setVcvars: ${{ parameters.SetVcvars }}
-
-    - task: CmdLine@1
-      displayName: 'Download test data and generate cmake config'
-      inputs:
-        filename: '$(Build.BinariesDirectory)\packages\python\python.exe'
-        arguments: '$(BuildCommand) --update --config Debug RelWithDebInfo'
-        workingDirectory: "$(Build.BinariesDirectory)"
-        
-    # Build Debug Mode
-    - ${{ if eq(parameters['DoDebugBuild'], 'true') }}:
-      - task: DotNetCoreCLI@2
-        displayName: 'Restore nuget packages for debug build'
-        inputs:
-          command: restore
-          projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-          configuration: 'Debug'          
-          arguments: '--configuration Debug -p:Platform="Any CPU" -p:OrtPackageId=${{ parameters.OrtPackageId}}'
-          workingDirectory: '$(Build.SourcesDirectory)\csharp'    
-      - task: VSBuild@1
-        displayName: 'Build Debug'
-        inputs:
-          solution: '$(Build.BinariesDirectory)\Debug\onnxruntime.sln'
-          platform: 'Win32'
-          configuration: 'Debug'
-          msbuildArgs: ${{ parameters.MsbuildArguments }}
-          msbuildArchitecture: $(BuildArch)
-          maximumCpuCount: true
-          logProjectEvents: true
-          workingFolder: '$(Build.BinariesDirectory)\Debug'
-          createLogFile: true
-
-      - task: BatchScript@1
-        displayName: 'Test Debug'
-        inputs:
-          filename: '$(Build.BinariesDirectory)\packages\python\python.exe'
-          arguments: '$(BuildCommand) --test --config Debug'
-          workingFolder: '$(Build.BinariesDirectory)'
-
-      - task: DotNetCoreCLI@2
-        displayName: 'Build C# Debug'
-        inputs:
-          command: build
-          projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-          configuration: 'Debug'          
-          arguments: '--configuration Debug -p:Platform="Any CPU" -p:OrtPackageId=${{ parameters.OrtPackageId}}'
-          workingDirectory: '$(Build.SourcesDirectory)\csharp'
-          
-
-    # Run test coverage report
-    - ${{ if eq(parameters['DoTestCoverage'], 'true') }}:
-      - template: windows-code-coverage-steps.yml
-        parameters:
-          OpenCppCoverageExe: '$(Build.BinariesDirectory)\OpenCppCoverage\OpenCppCoverage.exe'
-
-    # Build RelWithDebInfo -- this variable required to build C#
-    - script: |
-       @echo ##vso[task.setvariable variable=Configuration]RelWithDebInfo
-
-    - task: DotNetCoreCLI@2
-      displayName: 'Restore nuget packages for RelWithDebInfo build'
-      inputs:
-        command: restore
-        projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-        configuration: 'RelWithDebInfo'          
-        arguments: '--configuration RelWithDebInfo -p:Platform="Any CPU" -p:OrtPackageId=${{ parameters.OrtPackageId}}'
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-        
-    - task: VSBuild@1
-      displayName: 'Build RelWithDebInfo'
-      inputs:
-        solution: '$(Build.BinariesDirectory)\RelWithDebInfo\onnxruntime.sln'
-        platform: 'Win32'
-        configuration: 'RelWithDebInfo'
-        msbuildArgs: ${{ parameters.MsbuildArguments }}
-        msbuildArchitecture: $(BuildArch)
-        maximumCpuCount: true
-        logProjectEvents: true
-        workingFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
-        createLogFile: true
-
-    - task: BatchScript@1
-      displayName: 'Test RelWithDebInfo'
-      inputs:
-        filename: '$(Build.BinariesDirectory)\packages\python\python.exe'
-        arguments: '$(BuildCommand) --test --config RelWithDebInfo'
-        workingFolder: "$(Build.BinariesDirectory)"
-
-    - task: DotNetCoreCLI@2
-      displayName: 'Build C# RelWithDebInfo'
-      inputs:
-        command: build
-        projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-        configuration: 'RelWithDebInfo'          
-        arguments: '--configuration RelWithDebInfo -p:Platform="Any CPU" -p:OrtPackageId=${{ parameters.OrtPackageId}}'
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-
-    - task: DotNetCoreCLI@2
-      displayName: 'Test C# RelWithDebInfo'
-      inputs:
-        command: test
-        projects: '$(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.sln'
-        configuration: 'RelWithDebInfo'          
-        arguments: '--configuration RelWithDebInfo -p:Platform="Any CPU" -p:OrtPackageId=${{ parameters.OrtPackageId}}'
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-
-    - task: PublishTestResults@2
-      displayName: 'Publish unit test results'
-      inputs:
-        testResultsFiles: '**\*.results.xml'
-        searchFolder: '$(Build.BinariesDirectory)'
-        testRunTitle: 'Unit Test Run'
-      condition: succeededOrFailed()
-
-    # Nuget packaging if needed
-    - ${{ if eq(parameters['DoNugetPack'], 'true') }}:
-      - task: BatchScript@1
-        displayName: 'Setup VS2017 env vars'
-        inputs:
-          filename: 'C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvarsall.bat'
-          arguments: ${{ parameters.msbuildArchitecture }}
-          modifyEnvironment: true
-      # Esrp signing
-      - template: esrp_dll.yml
-        parameters:
-          FolderPath: '$(Build.BinariesDirectory)\RelWithDebInfo'
-          DisplayName: 'ESRP - Sign Native dlls'
-          DoEsrp: ${{ parameters.DoEsrp }}
-
-      - template: esrp_dll.yml
-        parameters:
-          FolderPath: '$(Build.SourcesDirectory)\csharp\src\Microsoft.ML.OnnxRuntime\bin\RelWithDebInfo'
-          DisplayName: 'ESRP - Sign C# dlls'
-          DoEsrp: ${{ parameters.DoEsrp }}
-
-      - script: |
-         ${{ parameters.NuPackScript }}
-        workingDirectory: '$(Build.SourcesDirectory)\csharp'
-        displayName: 'Create NuGet Package'
-
-      - task: PublishPipelineArtifact@0
-        displayName: 'Publish Pipeline Artifact: drop-nuget'
-        inputs:
-          artifactName: ${{ parameters.ArtifactName }}
-          targetPath: '$(Build.ArtifactStagingDirectory)'
-
-    # Compliance tasks require logs from Debug Build
-    - ${{ if eq(parameters['DoCompliance'], 'true') }}:
-      - template: compliance.yml
-
-    - task: ms.vss-governance-buildtask.governance-build-task-component-detection.ComponentGovernanceComponentDetection@0
-      displayName: 'Component Detection'
-      condition: succeeded()
-
-    - template: clean-agent-build-directory-step.yml
-
diff --git a/tools/ci_build/github/azure-pipelines/templates/windows-code-coverage-steps.yml b/tools/ci_build/github/azure-pipelines/templates/windows-code-coverage-steps.yml
index aff80c45b22b2..89edb0cc6e119 100644
--- a/tools/ci_build/github/azure-pipelines/templates/windows-code-coverage-steps.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/windows-code-coverage-steps.yml
@@ -12,7 +12,7 @@ steps:
   displayName: 'Run Test Coverage'
   inputs:
     filePath: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\run_OpenCppCoverage.ps1'
-    arguments: '-OpenCppCoverageExe:"${{parameters.OpenCppCoverageExe}}" -SourceRoot:"$(Build.SourcesDirectory)" -BuildRoot:"$(Build.BinariesDirectory)" -LocalBuild:$false'
+    arguments: '-OpenCppCoverageExe:"${{parameters.OpenCppCoverageExe}}" -SourceRoot:"$(Build.SourcesDirectory)" -BuildRoot:"$(Build.BinariesDirectory)" -LocalBuild:$true'
   continueOnError: ${{parameters.ContinueOnError}}  
 
 - task: PublishCodeCoverageResults@1
@@ -30,11 +30,11 @@ steps:
     arguments: 'install -q --insecure -y mysql-connector-python'
   timeoutInMinutes: 10
 
-- task: BatchScript@1
-  displayName: 'Post code coverage data to the Dashboard'
+- task: PythonScript@0
+  displayName: 'Generate cmake config'
   inputs:
-    filename: '$(Build.BinariesDirectory)\packages\python\python.exe'
-    arguments: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\post_code_coverage_to_dashboard.py --commit_hash=${{parameters.GitCommitHash}} --report_file="$(Build.BinariesDirectory)/Debug/Debug/cobertura.xml" --report_url="https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=$(Build.BuildId)"'
+    scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\post_code_coverage_to_dashboard.py'
+    arguments: --commit_hash=${{parameters.GitCommitHash}} --report_file="$(Build.BinariesDirectory)/Debug/Debug/cobertura.xml" --report_url="https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=$(Build.BuildId)"
     workingFolder: "$(Build.BinariesDirectory)"
   continueOnError: ${{parameters.ContinueOnError}}
   condition: ${{parameters.PostToDashBoard}}
diff --git a/tools/ci_build/github/windows/run_OpenCppCoverage.ps1 b/tools/ci_build/github/windows/run_OpenCppCoverage.ps1
index 05dbe23a1e8e2..8cb29607f69ca 100644
--- a/tools/ci_build/github/windows/run_OpenCppCoverage.ps1
+++ b/tools/ci_build/github/windows/run_OpenCppCoverage.ps1
@@ -56,21 +56,21 @@ $outputDir = Join-Path $buildDir "OpenCppCoverageResults"
 $modelDir = Join-Path $BuildRoot "models" 
 
 
-# ONNX test runner tests. 
+Write-Host "ONNX test runner tests"
 $onnx_test_runner = Join-Path $buildDir "onnx_test_runner.exe" 
 RunTest $onnx_test_runner ($modelDir) ("binary:"  + (Join-Path $buildDir "onnx_test_runner.cov"))
 
 
-# C-API/Shared-lib test
+Write-Host "C-API/Shared-lib test"
 $shared_lib_test = Join-Path $buildDir "onnxruntime_shared_lib_test.exe"
 RunTest $shared_lib_test @() ("binary:" + (Join-Path $buildDir "onnxruntime_shared_lib_test.cov"))
 
 
-# MLAS test
+Write-Host "MLAS test"
 $mlas_test = Join-Path $buildDir "onnxruntime_mlas_test.exe"
 RunTest $mlas_test @() ("binary:" + (Join-Path $buildDir "onnxruntime_mlas_test.cov"))
 
-# Lotus unit tests
+Write-Host "Lotus unit tests"
 # need to copy the tvm.dll, since it is not in the buildDir path
 if (Test-Path -Path $BuildRoot\Debug\external\tvm\Debug\tvm.dll -PathType Leaf) {
     Copy-Item -Path $BuildRoot\Debug\external\tvm\Debug\tvm.dll -Destination $buildDir

From da03ed447350314f662ee06d7829f2591ff25bef Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Wed, 13 Nov 2019 15:59:16 -0800
Subject: [PATCH 02/75] Tiny fix to codegen

---
 .../codegen/passes/op_ir_creator/tvm_ir_builder.cc    |  2 +-
 .../codegen/passes/weight_layout/weight_layout.cc     | 11 +++++------
 onnxruntime/test/tvm/tvm_basic_test.cc                |  2 +-
 onnxruntime/test/tvm/tvm_demo/demo_compiler.cc        |  2 +-
 4 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/onnxruntime/core/codegen/passes/op_ir_creator/tvm_ir_builder.cc b/onnxruntime/core/codegen/passes/op_ir_creator/tvm_ir_builder.cc
index a50d1ee3f3839..2fe892797886a 100644
--- a/onnxruntime/core/codegen/passes/op_ir_creator/tvm_ir_builder.cc
+++ b/onnxruntime/core/codegen/passes/op_ir_creator/tvm_ir_builder.cc
@@ -70,7 +70,7 @@ Status TVMIRBuilder::Evaluate(
 
 // BEGIN: Generic IR creator classes
 #define ADD_OP_ITEM(name) \
-  op_ir_registry->Register(std::move(onnxruntime::make_unique<GENERIC_OP_IR_CREATOR_CLASS(name)>()));
+  op_ir_registry->Register(onnxruntime::make_unique<GENERIC_OP_IR_CREATOR_CLASS(name)>());
 
 #define BINARY_OP(name) ADD_OP_ITEM(name)
 #define BINARY_CMP_OP(name) ADD_OP_ITEM(name)
diff --git a/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc b/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc
index 207b5c6efc5e1..7cbb5692f7096 100644
--- a/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc
+++ b/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc
@@ -31,12 +31,11 @@ const std::string WeightLayout::GetKey(
     ONNX_NAMESPACE::TensorProto_DataType proto_type,
     int input_dim,
     float pad_zero) {
-  std::string key = name;
-  key += "_type_" + std::to_string(static_cast<int>(proto_type));
-  key += "_dim_" + input_dim;
-  key += "_pad_zero_" + std::to_string(pad_zero);
-  key = NormalizeCppName(key);
-  return key;
+  std::ostringstream key(name);
+  key << "_type_" << static_cast<int>(proto_type);
+  key << "_dim_" << input_dim;
+  key << "_pad_zero_" << pad_zero;
+  return NormalizeCppName(key.str());
 }
 
 WeightLayout::WeightLayout(
diff --git a/onnxruntime/test/tvm/tvm_basic_test.cc b/onnxruntime/test/tvm/tvm_basic_test.cc
index 0b9d338d0d6e7..70399f2f03679 100644
--- a/onnxruntime/test/tvm/tvm_basic_test.cc
+++ b/onnxruntime/test/tvm/tvm_basic_test.cc
@@ -150,7 +150,7 @@ class FuseExecutionProviderX : public CPUExecutionProvider {
             onnxruntime::make_unique<ComputeCapability>(std::move(sub_graph)));
       }
     }
-    return std::move(result);
+    return result;
   }
 
   common::Status Compile(const std::vector<onnxruntime::Node*>& fused_nodes,
diff --git a/onnxruntime/test/tvm/tvm_demo/demo_compiler.cc b/onnxruntime/test/tvm/tvm_demo/demo_compiler.cc
index 886ecc9e1af31..15ca50f66664a 100644
--- a/onnxruntime/test/tvm/tvm_demo/demo_compiler.cc
+++ b/onnxruntime/test/tvm/tvm_demo/demo_compiler.cc
@@ -110,7 +110,7 @@ bool TVM_SCHEDULER_CLASS(AlwaysInline, DemoTVM)::Evaluate(
 // Register the always inline Scheduler to sched_registry
 static void RegisterAlwaysInlineScheduler(tvm_codegen::TVMScheduleRegistry* sched_registry) {
   sched_registry->Register(
-      std::move(onnxruntime::make_unique<TVM_SCHEDULER_CLASS(AlwaysInline, DemoTVM)>()));
+      onnxruntime::make_unique<TVM_SCHEDULER_CLASS(AlwaysInline, DemoTVM)>());
 }
 
 // Declare a schedule dispatcher that always dispatches the always inline Scheduler

From 4b900dc585045601f04326bb0ea5e4b565cd6241 Mon Sep 17 00:00:00 2001
From: KeDengMS <kedeng@microsoft.com>
Date: Fri, 20 Dec 2019 17:52:41 -0800
Subject: [PATCH 03/75] Simplify cache implementation and avoid static
 variables that may carry over between models

---
 onnxruntime/core/codegen/mti/mti_tvm_utils.cc |   5 +-
 onnxruntime/core/codegen/mti/mti_tvm_utils.h  |   2 +-
 .../passes/weight_layout/weight_layout.cc     |   4 +-
 .../nuphar/common/nuphar_tvm_utils.cc         | 155 ++++++++----------
 .../nuphar/common/nuphar_tvm_utils.h          |  11 +-
 .../nuphar/compiler/nuphar_codegen_ctx.cc     |  11 +-
 .../nuphar/compiler/nuphar_compiler.cc        |   9 +-
 .../python/onnxruntime_test_python_nuphar.py  |  51 +++---
 8 files changed, 121 insertions(+), 127 deletions(-)

diff --git a/onnxruntime/core/codegen/mti/mti_tvm_utils.cc b/onnxruntime/core/codegen/mti/mti_tvm_utils.cc
index 562834bf9580f..0afc25d627cd5 100644
--- a/onnxruntime/core/codegen/mti/mti_tvm_utils.cc
+++ b/onnxruntime/core/codegen/mti/mti_tvm_utils.cc
@@ -113,18 +113,17 @@ tvm::Tensor Promote(const tvm::Expr& expr, const tvm::Array<tvm::Expr>& shape, c
       name);
 }
 
-void DumpTVMModuleToFile(const std::string& filename_prefix, tvm::runtime::Module& module) {
+void DumpTVMModuleToFile(const std::string& filename, tvm::runtime::Module& module) {
   const codegen::CodeGenSettings& settings = codegen::CodeGenSettings::Instance();
   if (!settings.HasOption(codegen::CodeGenSettings::kCodeGenDumpModule))
     return;
 
-  static int dump_module_cnt = 0;
   // ISSUE: note that all option values are converted to lower case. It doesn't cause
   // any issue currently, because all supported formats (i.e. file exts) are of lower case.
   // Just keep in mind that we might have issue if somehow we started to support dump
   // formats with upper case, although it's quite unlikely.
   std::string format = settings.GetOptionValue(codegen::CodeGenSettings::kCodeGenDumpModule);
-  std::string module_filename = filename_prefix + "_" + std::to_string(dump_module_cnt++) + "." + format;
+  std::string module_filename = filename + "." + format;
   module->SaveToFile(module_filename, format);
 }
 
diff --git a/onnxruntime/core/codegen/mti/mti_tvm_utils.h b/onnxruntime/core/codegen/mti/mti_tvm_utils.h
index 709269aacff04..487c6415a7e7d 100644
--- a/onnxruntime/core/codegen/mti/mti_tvm_utils.h
+++ b/onnxruntime/core/codegen/mti/mti_tvm_utils.h
@@ -51,7 +51,7 @@ tvm::Tensor Promote(const tvm::Expr& expr,
 
 tvm::Tensor MakeZeroTensor(const tvm::Array<tvm::Expr>& shape, HalideIR::Type type, const std::string& name);
 
-void DumpTVMModuleToFile(const std::string& filename_prefix, tvm::runtime::Module& module);
+void DumpTVMModuleToFile(const std::string& filename, tvm::runtime::Module& module);
 
 bool BroadcastDim(const tvm::Array<tvm::Expr>& shape, size_t i, size_t output_rank, tvm::Expr& dim);
 
diff --git a/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc b/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc
index 7cbb5692f7096..ab3e647fd284a 100644
--- a/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc
+++ b/onnxruntime/core/codegen/passes/weight_layout/weight_layout.cc
@@ -31,8 +31,8 @@ const std::string WeightLayout::GetKey(
     ONNX_NAMESPACE::TensorProto_DataType proto_type,
     int input_dim,
     float pad_zero) {
-  std::ostringstream key(name);
-  key << "_type_" << static_cast<int>(proto_type);
+  std::ostringstream key;
+  key << name << "_type_" << static_cast<int>(proto_type);
   key << "_dim_" << input_dim;
   key << "_pad_zero_" << pad_zero;
   return NormalizeCppName(key.str());
diff --git a/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.cc b/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.cc
index f0fcf465580bd..7e27842f1ee83 100644
--- a/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.cc
+++ b/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.cc
@@ -15,7 +15,7 @@
 #include "gsl/gsl"
 #include <topi/detail/extern.h>
 #include <tvm/ir_pass.h>
-#define _SILENCE_EXPERIMENTAL_FILESYSTEM_DEPRECATION_WARNING // required by VS 2019
+#define _SILENCE_EXPERIMENTAL_FILESYSTEM_DEPRECATION_WARNING  // required by VS 2019
 #include <experimental/filesystem>
 #undef _SILENCE_EXPERIMENTAL_FILESYSTEM_DEPRECATION_WARNING
 #include <atomic>
@@ -102,117 +102,102 @@ static void ParseVersion(const char* version, int* major, int* minor, int* patch
   *patch = ver_num_fn(val);
 }
 
-static void VerifyCacheVersion(const std::string& so_path) {
-  static std::atomic<bool> cache_version_checked{false};
-  static std::mutex cache_version_mutex;
-
-  // make sure we only check cache version once
-  if (!cache_version_checked.load(std::memory_order::memory_order_acquire)) {
-    std::lock_guard<std::mutex> lock(cache_version_mutex);
-    if (!cache_version_checked.load(std::memory_order::memory_order_acquire)) {
-      cache_version_checked.store(true, std::memory_order::memory_order_release);
-      // ensure we have _ORTInternal_GetCacheVersion_ function
-      void* f = GetFuncFromLibrary(so_path, "_ORTInternal_GetCacheVersion", /*throw_if_not_found*/ true);
-      ORT_ENFORCE(f, "NULL library function pointer!");
-
-      typedef const char* (*GetVersionFunc)();
-      GetVersionFunc func = reinterpret_cast<GetVersionFunc>(f);
-      const char* cache_version = func();
-      ORT_ENFORCE(cache_version, "Null cache version string!");
-      int cur_major, cur_minor, cur_patch;
-      ParseVersion(__NUPHAR_CACHE_VERSION__, &cur_major, &cur_minor, &cur_patch);
-      int cache_major, cache_minor, cache_patch;
-      ParseVersion(cache_version, &cache_major, &cache_minor, &cache_patch);
-
-      // make version check strict until we have thorough design for compatibility
-      ORT_ENFORCE((cur_major == cache_major) && (cur_minor == cache_minor),
-                  "Current nuphar runtime version (", __NUPHAR_CACHE_VERSION__,
-                  ") doesn't match cached dll version (", cache_version, ")");
-
-      cache_version_checked = true;
-    }
+static bool VerifyCacheVersion(const std::string& so_path) {
+  // ensure we have _ORTInternal_GetCacheVersion_ function
+  void* f = GetFuncFromLibrary(so_path, "_ORTInternal_GetCacheVersion", /*throw_if_not_found*/ true);
+  if (f == nullptr)
+    return false;
+
+  typedef const char* (*GetVersionFunc)();
+  GetVersionFunc func = reinterpret_cast<GetVersionFunc>(f);
+  const char* cache_version = func();
+  ORT_ENFORCE(cache_version, "Null cache version string!");
+  int cur_major, cur_minor, cur_patch;
+  ParseVersion(__NUPHAR_CACHE_VERSION__, &cur_major, &cur_minor, &cur_patch);
+  int cache_major, cache_minor, cache_patch;
+  ParseVersion(cache_version, &cache_major, &cache_minor, &cache_patch);
+
+  // make version check strict until we have thorough design for compatibility
+  bool version_match = (cur_major == cache_major) && (cur_minor == cache_minor);
+  if (!version_match) {
+    LOGS_DEFAULT(CODEGEN_SETTINGS_LOG_LEVEL) << "Current nuphar runtime version (" << __NUPHAR_CACHE_VERSION__ << ") doesn't match cached dll version (" << cache_version << ")";
   }
+  return version_match;
 }
 
-static bool disable_caching_due_to_checksum_failure = false;
-
 static bool VerifyTVMModuleChecksum(const std::string& so_path) {
-  static std::string last_so_path;
-  static bool last_checksum_validated = false;
-  static std::mutex checksum_mutex;
-  if (last_so_path != so_path) {
-    std::lock_guard<std::mutex> lock(checksum_mutex);
-    if (last_so_path != so_path) {
-      disable_caching_due_to_checksum_failure = false;  // reset disabled caching for a new file
-      last_so_path = so_path;
-      void* f = GetFuncFromLibrary(so_path, "_ORTInternal_GetCheckSum", /*throw_if_not_found*/ false);
-      if (f) {
-        typedef void (*GetChecksumFunc)(const char*&, size_t&);
-        GetChecksumFunc func = reinterpret_cast<GetChecksumFunc>(f);
-        const char* model_checksum;
-        size_t model_checksum_len;
-        func(model_checksum,
-             model_checksum_len);
-
-        codegen::CodeGenSettings& setting = codegen::CodeGenSettings::Instance();
-        // When checksum is expected by dll/so, user must set environment variable
-        // NUPHAR_CACHE_MODEL_CHECKSUM from md5 digest of running model.
-        // User may choose to run with base model or simplified mode and any match
-        // would be regarded as validated.
-        // Note that checksum validation here is not designed as a security measurement,
-        // so checksum compute is not done inside ORT.
-        last_checksum_validated =
-            setting.OptionMatches(
-                kNupharCacheModelChecksum,
-                std::string(model_checksum, model_checksum_len));
-
-        if (!last_checksum_validated) {
-          LOGS_DEFAULT(CODEGEN_SETTINGS_LOG_LEVEL) << "Cache checksum validation failed, using JIT...";
-          disable_caching_due_to_checksum_failure = true;
-        }
-      } else {
-        // do not validate checksum if dll didn't require it (usually during debugging)
-        // TODO: force checksum validation in final release
-        last_checksum_validated = true;
+  void* f = GetFuncFromLibrary(so_path, "_ORTInternal_GetCheckSum", /*throw_if_not_found*/ false);
+  if (f) {
+    typedef void (*GetChecksumFunc)(const char*&, size_t&);
+    GetChecksumFunc func = reinterpret_cast<GetChecksumFunc>(f);
+    const char* model_checksum;
+    size_t model_checksum_len;
+    func(model_checksum,
+         model_checksum_len);
+
+    codegen::CodeGenSettings& setting = codegen::CodeGenSettings::Instance();
+    // When checksum is expected by dll/so, user must set environment variable
+    // NUPHAR_CACHE_MODEL_CHECKSUM from md5 digest of running model.
+    // User may choose to run with base model or simplified mode and any match
+    // would be regarded as validated.
+    // Note that checksum validation here is not designed as a security measurement,
+    // so checksum compute is not done inside ORT.
+    if (setting.OptionMatches(
+            kNupharCacheModelChecksum,
+            std::string(model_checksum, model_checksum_len))) {
+      return true;
+    } else {
+      static std::mutex warn_mutex;
+      std::lock_guard<std::mutex> warn_lock(warn_mutex);
+      static std::string last_warned_so_path;
+      if (last_warned_so_path != so_path) {
+        // warning only once for each so_path
+        LOGS_DEFAULT(CODEGEN_SETTINGS_LOG_LEVEL) << "Cache checksum validation failed, using JIT...";
       }
+      return false;
     }
+  } else {
+    // do not validate checksum if dll didn't require it (usually during debugging)
+    // TODO: force checksum validation in final release
+    return true;
   }
-  return last_checksum_validated;
 }
 
-tvm::runtime::PackedFunc LoadTVMPackedFuncFromCache(const std::string& func_name) {
+CacheStatus LoadTVMPackedFuncFromCache(const std::string& func_name, tvm::runtime::PackedFunc& func) {
   std::string so_path;
-  if (!GetCacheSoFilePath(so_path))
-    return nullptr;
+  if (!GetCacheSoFilePath(so_path)) {
+    if (codegen::CodeGenSettings::Instance().HasOption(kNupharCachePath)) {
+      return CacheStatus::Missing;
+    } else {
+      return CacheStatus::NotInUse;
+    }
+  }
 
-  VerifyCacheVersion(so_path);
+  if (!VerifyCacheVersion(so_path)) {
+    return CacheStatus::Mismatch;
+  }
 
   if (!VerifyTVMModuleChecksum(so_path))
-    return nullptr;
+    return CacheStatus::Mismatch;
 
   tvm::runtime::Module module = tvm::runtime::Module::LoadFromFile(so_path);
-  tvm::runtime::PackedFunc func = module.GetFunction(func_name);
+  func = module.GetFunction(func_name);
   if (func == nullptr) {
     LOGS_DEFAULT(CODEGEN_SETTINGS_LOG_LEVEL) << "Cannot find " << func_name << " in cache, using JIT...";
+    return CacheStatus::Missing;
   }
-  return func;
+  return CacheStatus::Found;
 }
 
 void SaveTVMModuleToCache(const std::string& filename, tvm::runtime::Module& module) {
   fs::path path;
 
-  if (disable_caching_due_to_checksum_failure)
-    return;
-
   static std::mutex save_cache_mutex;
-  static std::unordered_set<std::string> existing_files;
   std::lock_guard<std::mutex> lock(save_cache_mutex);
-  if (existing_files.count(filename) == 0 &&
-      GetOrCreateTVMModuleCacheDirectory(path, /*create*/ true)) {
-    existing_files.insert(filename);
+  if (GetOrCreateTVMModuleCacheDirectory(path, /*create*/ true)) {
     path.append(filename + ".o");
     if (fs::exists(path)) {
-      LOGS_DEFAULT(CODEGEN_SETTINGS_LOG_LEVEL) << "Object file " << path << " already exists, skip saving...";
+      //LOGS_DEFAULT(CODEGEN_SETTINGS_LOG_LEVEL) << "Object file " << path << " already exists, skip saving...";
       return;
     }
     module->SaveToFile(path.string(), "o");
diff --git a/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.h b/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.h
index fe9162911ce52..a9f75bdf17221 100644
--- a/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.h
+++ b/onnxruntime/core/providers/nuphar/common/nuphar_tvm_utils.h
@@ -19,8 +19,15 @@ struct NupharSubgraphUnit;  //forward
 // Helper functions to create or load from offline cached dll
 // note after saving to obj file, we need to use tvm Python to create dll
 // using script at onnxruntime/core/codegen/mti/scripts/create_shared.py
-tvm::runtime::PackedFunc
-LoadTVMPackedFuncFromCache(const std::string& func_name);
+
+enum class CacheStatus {
+  NotInUse,
+  Mismatch,
+  Missing,
+  Found,
+};
+
+CacheStatus LoadTVMPackedFuncFromCache(const std::string& func_name, tvm::runtime::PackedFunc& func);
 void SaveTVMModuleToCache(const std::string& filename, tvm::runtime::Module& module);
 
 std::string GetPackedFuncName(const nuphar::NupharSubgraphUnit& subgraph, const CodeGenTarget& codegen_target, int64_t parallel_min_workloads);
diff --git a/onnxruntime/core/providers/nuphar/compiler/nuphar_codegen_ctx.cc b/onnxruntime/core/providers/nuphar/compiler/nuphar_codegen_ctx.cc
index 08031691c5ee7..f96913f2f9c38 100644
--- a/onnxruntime/core/providers/nuphar/compiler/nuphar_codegen_ctx.cc
+++ b/onnxruntime/core/providers/nuphar/compiler/nuphar_codegen_ctx.cc
@@ -56,13 +56,16 @@ static tvm::runtime::PackedFunc LowerLayoutFunc(const tvm_codegen::WeightLayout*
 
   std::string func_name = layout->Name() + "_marshall";
 
-  tvm::runtime::PackedFunc cached_func = nuphar::LoadTVMPackedFuncFromCache(func_name);
-
-  if (cached_func == nullptr) {
+  tvm::runtime::PackedFunc cached_func;
+  auto cache_status = nuphar::LoadTVMPackedFuncFromCache(func_name, cached_func);
+  if (cache_status != nuphar::CacheStatus::Found) {
+    ORT_ENFORCE(cached_func == nullptr);
     auto lowered = tvm::lower(S, {inputs[0], outputs[0]}, func_name, {}, config);
     auto module = tvm::build(lowered, tvm::target::llvm(), tvm::Target(), config);
     tvm_codegen::DumpTVMModuleToFile(func_name, module);
-    nuphar::SaveTVMModuleToCache(func_name, module);
+    if (cache_status == nuphar::CacheStatus::Missing) {
+      nuphar::SaveTVMModuleToCache(func_name, module);
+    }
     cached_func = module.GetFunction(func_name);
   }
   return cached_func;
diff --git a/onnxruntime/core/providers/nuphar/compiler/nuphar_compiler.cc b/onnxruntime/core/providers/nuphar/compiler/nuphar_compiler.cc
index 9b41ac88cf4a0..4b648e0e386a7 100644
--- a/onnxruntime/core/providers/nuphar/compiler/nuphar_compiler.cc
+++ b/onnxruntime/core/providers/nuphar/compiler/nuphar_compiler.cc
@@ -151,8 +151,9 @@ tvm::runtime::PackedFunc NupharCompiler::GetLoweredPackedFunc(
   // JIT-caching and AOT are mutual exclusive.
   // Change it by not always saving a compiled func unless it is in JIT-Caching model.
   // In AOT, there should be another member func explicitly loading
-  tvm::runtime::PackedFunc cached_func = nuphar::LoadTVMPackedFuncFromCache(func_name);
-  if (cached_func == nullptr) {
+  tvm::runtime::PackedFunc cached_func;
+  auto cache_status = nuphar::LoadTVMPackedFuncFromCache(func_name, cached_func);
+  if (cache_status != nuphar::CacheStatus::Found) {
     codegen::CodeGenSettings& settings = codegen::CodeGenSettings::Instance();
 
     if (settings.HasOption(kNupharCacheForceNoJIT)) {
@@ -180,7 +181,9 @@ tvm::runtime::PackedFunc NupharCompiler::GetLoweredPackedFunc(
 
     tvm::runtime::Module module = tvm::build(lowered, tvm_target, tvm_host_target, config);
     tvm_codegen::DumpTVMModuleToFile(func_name, module);
-    nuphar::SaveTVMModuleToCache(func_name, module);
+    if (cache_status == nuphar::CacheStatus::Missing) {
+      nuphar::SaveTVMModuleToCache(func_name, module);
+    }
     cached_func = module.GetFunction(func_name);
   }
 
diff --git a/onnxruntime/test/python/onnxruntime_test_python_nuphar.py b/onnxruntime/test/python/onnxruntime_test_python_nuphar.py
index b496a2b93e6a6..f9387ffb73c17 100644
--- a/onnxruntime/test/python/onnxruntime_test_python_nuphar.py
+++ b/onnxruntime/test/python/onnxruntime_test_python_nuphar.py
@@ -41,11 +41,13 @@ def test_bidaf(self):
 
         # run onnx_test_runner to verify results
         # use -M to disable memory pattern
-        # use -j 1 -c 1 to run one model/session at a time when running multiple models
         onnx_test_runner = os.path.join(cwd, 'onnx_test_runner')
-        subprocess.run([onnx_test_runner, '-e', 'nuphar', '-M', '-c', '1', '-j', '1', '-n', 'bidaf', cwd], check=True, cwd=cwd)
+        subprocess.run([onnx_test_runner, '-e', 'nuphar', '-M', '-n', 'bidaf', cwd], check=True, cwd=cwd)
 
         # test AOT on the quantized model
+        if os.name not in ['nt', 'posix']:
+            return # don't run the rest of test if AOT is not supported
+
         cache_dir = os.path.join(cwd, 'nuphar_cache')
         if os.path.exists(cache_dir):
             shutil.rmtree(cache_dir)
@@ -57,33 +59,28 @@ def test_bidaf(self):
             tp = onnx.load_tensor(os.path.join(bidaf_dir, 'test_data_set_0', 'input_{}.pb'.format(i)))
             feed[tp.name] = numpy_helper.to_array(tp)
 
-        # force codegen_target to be avx
-        nuphar_settings = 'nuphar_codegen_target:avx'
-        onnxrt.capi._pybind_state.set_nuphar_settings(nuphar_settings)
-        sess = onnxrt.InferenceSession(bidaf_int8_scan_only_model)
-        assert 'NupharExecutionProvider' in sess.get_providers()
-        output = sess.run([], feed)
-
-        nuphar_settings = 'nuphar_cache_path:{}'.format(cache_dir)
-        onnxrt.capi._pybind_state.set_nuphar_settings(nuphar_settings)
-        sess = onnxrt.InferenceSession(bidaf_int8_scan_only_model) # JIT cache happens when initializing session
-        assert 'NupharExecutionProvider' in sess.get_providers()
-        output = sess.run([], feed)
-
-        cache_dir_content = os.listdir(cache_dir)
-        assert len(cache_dir_content) == 1
-        cache_versioned_dir = os.path.join(cache_dir, cache_dir_content[0])
-        so_name = 'bidaf.so'
-        if os.name in ['nt', 'posix'] : # Windows or Linux
+        for model in [bidaf_opt_scan_model, bidaf_int8_scan_only_model]:
+            nuphar_settings = 'nuphar_cache_path:{}'.format(cache_dir)
+            for isa in ['avx', 'avx2', 'avx512']:
+                onnxrt.capi._pybind_state.set_nuphar_settings(nuphar_settings + ', nuphar_codegen_target:' + isa)
+                sess = onnxrt.InferenceSession(model) # JIT cache happens when initializing session
+
+            cache_dir_content = os.listdir(cache_dir)
+            assert len(cache_dir_content) == 1
+            cache_versioned_dir = os.path.join(cache_dir, cache_dir_content[0])
+            so_name = os.path.basename(model) + '.so'
             subprocess.run([sys.executable, '-m', 'onnxruntime.nuphar.create_shared', '--input_dir', cache_versioned_dir, '--output_name', so_name], check=True)
-        else:
-            return # don't run the rest of test if AOT is not supported
 
-        nuphar_settings = 'nuphar_cache_path:{}, nuphar_cache_so_name:{}, nuphar_cache_force_no_jit:{}'.format(cache_dir, so_name, 'on')
-        onnxrt.capi._pybind_state.set_nuphar_settings(nuphar_settings)
-        sess = onnxrt.InferenceSession(bidaf_int8_scan_only_model) # JIT cache happens when initializing session
-        assert 'NupharExecutionProvider' in sess.get_providers()
-        sess.run([], feed)
+            nuphar_settings = 'nuphar_cache_path:{}, nuphar_cache_so_name:{}, nuphar_cache_force_no_jit:{}'.format(cache_dir, so_name, 'on')
+            onnxrt.capi._pybind_state.set_nuphar_settings(nuphar_settings)
+            sess = onnxrt.InferenceSession(model)
+            sess.run([], feed)
+
+            # test avx
+            nuphar_settings = 'nuphar_cache_path:{}, nuphar_cache_so_name:{}, nuphar_cache_force_no_jit:{}, nuphar_codegen_target:{}'.format(cache_dir, so_name, 'on', 'avx')
+            onnxrt.capi._pybind_state.set_nuphar_settings(nuphar_settings)
+            sess = onnxrt.InferenceSession(model)
+            sess.run([], feed)
 
 
     def test_bert_squad(self):

From bbe32f0990c4bbc8b0dc5a2221ee24b705e2a4ee Mon Sep 17 00:00:00 2001
From: Jeff <38966965+jeffbloo@users.noreply.github.com>
Date: Fri, 20 Dec 2019 23:28:15 -0800
Subject: [PATCH 04/75] Extend DML kernels (#2641)

* Additional DML operators

* Check unsupported attributes and inputs

* Address PR comments

* Add kernel capability function used for partitioning, and re-enable stride-based int64 support based on value range

* Fix test failures

* Build fix

* PR comments
---
 .../inc/IWinmlExecutionProvider.h             |  14 +-
 .../src/AbiCustomRegistry.cpp                 | 109 ++++++----
 .../src/AbiCustomRegistry.h                   |  10 +-
 .../src/ExecutionProvider.cpp                 |  15 +-
 .../src/ExecutionProvider.h                   |   2 +-
 .../src/GraphDescBuilder.cpp                  |   4 +-
 .../src/GraphDescBuilder.h                    |   4 +-
 .../src/GraphPartitioner.cpp                  | 128 +++++++++---
 .../src/GraphPartitioner.h                    |   4 +-
 .../src/MLOperatorAuthorImpl.cpp              |   7 +
 .../src/MLOperatorAuthorImpl.h                |  15 ++
 .../src/Operators/DmlOperatorConvolution.cpp  |  32 +--
 .../src/Operators/DmlOperatorCopy.cpp         |   5 +-
 .../src/Operators/DmlOperatorPooling.cpp      |  44 ++++
 .../src/Operators/DmlOperatorSlice.cpp        |  34 ++-
 .../src/Operators/OperatorRegistration.cpp    |  56 +++--
 .../src/Operators/OperatorRegistration.h      |   1 +
 .../dml/OperatorAuthorHelper/Attributes.h     |   2 +
 .../MLOperatorAuthorHelper.h                  |  24 +++
 .../MLOperatorAuthorPrivate.h                 |  70 +++++--
 .../OperatorAuthorHelper/OperatorHelper.cpp   |  53 +----
 .../dml/OperatorAuthorHelper/OperatorHelper.h | 196 ++++++++++++++++--
 .../OperatorRegistration.h                    |   2 +
 onnxruntime/test/onnx/main.cc                 |   1 -
 24 files changed, 621 insertions(+), 211 deletions(-)

diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/inc/IWinmlExecutionProvider.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/inc/IWinmlExecutionProvider.h
index 37b88e891bafc..c34dfbc2d93d6 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/inc/IWinmlExecutionProvider.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/inc/IWinmlExecutionProvider.h
@@ -94,14 +94,22 @@ namespace winrt::Windows::AI::MachineLearning::implementation
         const void* executionHandle,
         DmlGraphNodeCreateInfo* graphNodeCreateInfo
         )>;
-
+    
     struct GraphNodeFactoryRegistration
     {
         GraphNodeFactory factory;
         std::optional<uint32_t> requiredInputCount;
-        std::vector<uint32_t> requiredConstantCpuInputs;
         bool requiresFloatFormatsExceptConstInputs = false;
     };
 
-    using GraphNodeFactoryMap = std::unordered_map<onnxruntime::KernelDef*, std::shared_ptr<GraphNodeFactoryRegistration>>;
+    using KernelSupportQuery = std::function<bool(const onnxruntime::Node& node)>;
+
+    struct InternalRegistrationInfo
+    {
+        std::vector<uint32_t> requiredConstantCpuInputs;
+        std::optional<GraphNodeFactoryRegistration> graphNodeFactoryRegistration;
+        KernelSupportQuery supportQuery;
+    };
+
+    using InternalRegistrationInfoMap = std::unordered_map<onnxruntime::KernelDef*, std::shared_ptr<InternalRegistrationInfo>>;
 }
\ No newline at end of file
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.cpp
index 5365d10c9f03e..76b08ca895095 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.cpp
@@ -9,7 +9,7 @@ namespace winrt::Windows::AI::MachineLearning::implementation
 
 AbiCustomRegistry::AbiCustomRegistry() : 
     m_kernelRegistry(std::make_shared<onnxruntime::CustomRegistry>()),
-    m_graphNodeFactoryMap(std::make_shared<GraphNodeFactoryMap>())
+    m_internalRegInfoMap(std::make_shared<InternalRegistrationInfoMap>())
 {
 }
 
@@ -321,13 +321,14 @@ HRESULT STDMETHODCALLTYPE AbiCustomRegistry::RegisterOperatorKernel(
     IMLOperatorKernelFactory* operatorKernelFactory,
     _In_opt_ IMLOperatorShapeInferrer* shapeInferrer) const noexcept 
 {
-    return RegisterOperatorKernel(opKernel, operatorKernelFactory, shapeInferrer, false, false, false);
+    return RegisterOperatorKernel(opKernel, operatorKernelFactory, shapeInferrer, nullptr, false, false, false);
 }
 
 HRESULT STDMETHODCALLTYPE AbiCustomRegistry::RegisterOperatorKernel(
     const MLOperatorKernelDescription* opKernel,
     IMLOperatorKernelFactory* operatorKernelFactory,
     _In_opt_ IMLOperatorShapeInferrer* shapeInferrer,
+    _In_opt_ IMLOperatorSupportQueryPrivate* supportQuery,
     bool isInternalOperator,
     bool canAliasFirstInput,
     bool supportsGraph,
@@ -449,63 +450,91 @@ HRESULT STDMETHODCALLTYPE AbiCustomRegistry::RegisterOperatorKernel(
         };
 
     onnxruntime::KernelCreateInfo create_info(builder.Build(), lotusKernelCreateFn);
+    onnxruntime::KernelDef* kernelDef = create_info.kernel_def.get();
 
-    if (supportsGraph)
+    if (isInternalOperator)
     {
+        auto regInfo = std::make_shared<InternalRegistrationInfo>();
+        regInfo->requiredConstantCpuInputs = constantCpuInputCapture;
+
         // Only internal operators support usage in DML graphs
-        if (!isInternalOperator)
+        if (supportsGraph)
         {
-            THROW_HR(E_INVALIDARG);
+            GraphNodeFactoryRegistration graphReg;
+            graphReg.factory = 
+                [kernelFactoryCapture,
+                requiresInputShapesAtCreation,
+                requiresOutputShapesAtCreation,
+                shapeInferrerCapture,
+                defaultAttributesCapture,
+                constantCpuInputCapture](const onnxruntime::Node& node, MLOperatorTensorGetter& constantInputGetter, const void* executionHandle, DmlGraphNodeCreateInfo* graphNodeCreateInfo)
+                {
+                    onnxruntime::ProtoHelperNodeContext nodeContext(node);
+                    onnxruntime::OpNodeProtoHelper<onnxruntime::ProtoHelperNodeContext> protoHelper(&nodeContext);
+                
+                    // Use the same list of required constant inputs for the shape inferrer and the kernel.
+                    EdgeShapes outputShapes;
+                    InferAndVerifyOutputSizes(node, &defaultAttributesCapture, shapeInferrerCapture.Get(), constantCpuInputCapture, constantInputGetter, nullptr, outputShapes);
+
+                    // Create the kernel while allowing input shape and output shape queries according to options
+                    ComPtr<DmlGraphOpKernelInfoWrapper> kernelInfoWrapper = wil::MakeOrThrow<DmlGraphOpKernelInfoWrapper>(
+                            &protoHelper,
+                            executionHandle,
+                            true,
+                            &outputShapes,
+                            &defaultAttributesCapture,
+                            graphNodeCreateInfo,
+                            constantCpuInputCapture,
+                            constantInputGetter);
+
+                    Microsoft::WRL::ComPtr<IMLOperatorKernel> kernel;
+                    THROW_IF_FAILED(kernelFactoryCapture->CreateKernel(kernelInfoWrapper.Get(), kernel.GetAddressOf()));
+                    kernelInfoWrapper->Close();
+                };
+
+            if (requiredInputCountForGraph)
+            {
+                graphReg.requiredInputCount = *requiredInputCountForGraph;
+            }
+
+            graphReg.requiresFloatFormatsExceptConstInputs = requiresFloatFormatsForGraph;
+            regInfo->graphNodeFactoryRegistration = graphReg;
         }
 
-        auto registration = std::make_shared<GraphNodeFactoryRegistration>();
+        if (supportQuery)
+        {
+            ComPtr<IMLOperatorSupportQueryPrivate> supportQueryCapture = supportQuery;
 
-        registration->factory = 
-            [kernelFactoryCapture,
-            requiresInputShapesAtCreation,
-            requiresOutputShapesAtCreation,
-            shapeInferrerCapture,
-            defaultAttributesCapture,
-            constantCpuInputCapture](const onnxruntime::Node& node, MLOperatorTensorGetter& constantInputGetter, const void* executionHandle, DmlGraphNodeCreateInfo* graphNodeCreateInfo)
+            regInfo->supportQuery = [supportQueryCapture, defaultAttributesCapture](const onnxruntime::Node& node)
             {
                 onnxruntime::ProtoHelperNodeContext nodeContext(node);
                 onnxruntime::OpNodeProtoHelper<onnxruntime::ProtoHelperNodeContext> protoHelper(&nodeContext);
-                
-                // Use the same list of required constant inputs for the shape inferrer and the kernel.
-                EdgeShapes outputShapes;
-                InferAndVerifyOutputSizes(node, &defaultAttributesCapture, shapeInferrerCapture.Get(), constantCpuInputCapture, constantInputGetter, nullptr, outputShapes);
-
+                              
                 // Create the kernel while allowing input shape and output shape queries according to options
-                ComPtr<DmlGraphOpKernelInfoWrapper> kernelInfoWrapper = wil::MakeOrThrow<DmlGraphOpKernelInfoWrapper>(
+                ComPtr<MLSupportQueryContext> supportContext = wil::MakeOrThrow<MLSupportQueryContext>(
                         &protoHelper,
-                        executionHandle,
-                        true,
-                        &outputShapes,
-                        &defaultAttributesCapture,
-                        graphNodeCreateInfo,
-                        constantCpuInputCapture,
-                        constantInputGetter);
-
-                Microsoft::WRL::ComPtr<IMLOperatorKernel> kernel;
-                THROW_IF_FAILED(kernelFactoryCapture->CreateKernel(kernelInfoWrapper.Get(), kernel.GetAddressOf()));
-                kernelInfoWrapper->Close();
-            };
+                        &defaultAttributesCapture);
 
-        if (requiredInputCountForGraph)
-        {
-            registration->requiredInputCount = *requiredInputCountForGraph;
+                BOOL bSupported = FALSE;
+                THROW_IF_FAILED(supportQueryCapture->QuerySupport(supportContext.Get(), &bSupported));
+                return !!bSupported;
+            };
         }
 
-        registration->requiresFloatFormatsExceptConstInputs = requiresFloatFormatsForGraph;
-        registration->requiredConstantCpuInputs = constantCpuInputCapture;
-
-        onnxruntime::KernelDef* kernelDef = create_info.kernel_def.get();
         THROW_IF_NOT_OK(m_kernelRegistry->RegisterCustomKernel(create_info));
-        (*m_graphNodeFactoryMap)[kernelDef] = registration;
+        (*m_internalRegInfoMap)[kernelDef] = regInfo;
     }
     else
     {
-        // For backward compatibility, this does not propagate errors
+        // Currently unsupported for external operators
+        if (canAliasFirstInput || supportsGraph || requiredInputCountForGraph || 
+            requiresFloatFormatsForGraph || requiredConstantCpuInputs)
+        {
+            THROW_HR(E_INVALIDARG);
+        }
+
+        //
+        // For backward compatibility, this does not propagate errors for external operators
         m_kernelRegistry->RegisterCustomKernel(create_info);
     }
 
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.h
index d3b90538308c1..dee783cd55a7d 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/AbiCustomRegistry.h
@@ -36,6 +36,7 @@ class AbiCustomRegistry : public WRL::Base<IMLOperatorRegistry, IMLOperatorRegis
         const MLOperatorKernelDescription* operatorKernel,
         IMLOperatorKernelFactory* operatorKernelFactory,
         _In_opt_ IMLOperatorShapeInferrer* shapeInferrer,
+        _In_opt_ IMLOperatorSupportQueryPrivate* supportQuery,
         bool isInternalOperator,
         bool canAliasFirstInput,
         bool supportsGraph,
@@ -79,9 +80,9 @@ class AbiCustomRegistry : public WRL::Base<IMLOperatorRegistry, IMLOperatorRegis
         return m_kernelRegistry;
     }
 
-    std::shared_ptr<GraphNodeFactoryMap> GetGraphNodeFactoryMap() const
+    std::shared_ptr<InternalRegistrationInfoMap> GetInternalRegInfoMap() const
     {
-        return m_graphNodeFactoryMap;
+        return m_internalRegInfoMap;
     }
 
  private:
@@ -104,8 +105,9 @@ class AbiCustomRegistry : public WRL::Base<IMLOperatorRegistry, IMLOperatorRegis
     // (see LotusOpSchemaRegistry::GetSchemaAndHistory).
     mutable std::map<std::pair<int, int>, std::shared_ptr<onnxruntime::CustomRegistry>> m_customRegistryOpsetVerMap;
 
-    // Map between Lotus KernelDefs and graph node factories used for fusing nodes for graph compilation
-    mutable std::shared_ptr<GraphNodeFactoryMap> m_graphNodeFactoryMap;
+    // Map between Lotus KernelDefs and extended data used during partitioning
+    mutable std::shared_ptr<InternalRegistrationInfoMap> m_internalRegInfoMap;
+
 };
 
 }    // namespace winrt::Windows::AI::MachineLearning::implementation
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.cpp
index 5563d9574653f..9352bfdae724d 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.cpp
@@ -45,7 +45,7 @@ namespace Dml
 
     static void CreateDmlKernelRegistry(
         _Outptr_ std::shared_ptr<onnxruntime::KernelRegistry>* registry,
-        _Outptr_ std::shared_ptr<const GraphNodeFactoryMap>* graphNodeFactoryMap)
+        _Outptr_ std::shared_ptr<const InternalRegistrationInfoMap>* internalRegInfoMap)
     {
         ComPtr<AbiCustomRegistry> abiRegistry = wil::MakeOrThrow<AbiCustomRegistry>();
         Dml::RegisterDmlOperators(abiRegistry.Get());
@@ -54,7 +54,7 @@ namespace Dml
         
         auto customRegistry = *abiRegistry->GetRegistries().begin();
         *registry = customRegistry->GetKernelRegistry();
-        *graphNodeFactoryMap = abiRegistry->GetGraphNodeFactoryMap();
+        *internalRegInfoMap = abiRegistry->GetInternalRegInfoMap();
     }
 
     ExecutionProvider::ExecutionProvider(
@@ -176,7 +176,7 @@ namespace Dml
         m_cpuInputAllocator = std::make_shared<CPUAllocator>(OrtMemType::OrtMemTypeCPUInput);
         m_cpuOutputAllocator = std::make_shared<CPUAllocator>(OrtMemType::OrtMemTypeCPUOutput);
 		
-        CreateDmlKernelRegistry(&m_kernelRegistry, &m_graphNodeFactoryMap);
+        CreateDmlKernelRegistry(&m_kernelRegistry, &m_internalRegInfoMap);
     }
 
     HRESULT __stdcall ExecutionProviderImpl::GetD3DDevice(_COM_Outptr_ ID3D12Device** d3dDevice) const noexcept
@@ -493,7 +493,14 @@ namespace Dml
     {
         std::string partitionKernelPrefix = std::to_string(m_partitionKernelPrefixVal++) + "_";
         uint32_t deviceDataTypeMask = GetSuppportedDeviceDataTypeMask();
-        return PartitionGraph(graph, *m_graphNodeFactoryMap, registries, deviceDataTypeMask, m_kernelRegistry.get(), partitionKernelPrefix);
+
+        return PartitionGraph(graph, 
+            *m_internalRegInfoMap,
+            registries, 
+            deviceDataTypeMask, 
+            m_kernelRegistry.get(), 
+            partitionKernelPrefix
+        );
     }
 
     Status ExecutionProviderImpl::CopyTensor(const onnxruntime::Tensor& src, onnxruntime::Tensor& dst) const
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
index 39e6d2ba01329..da8b405221598 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
@@ -172,7 +172,7 @@ namespace Dml
         std::shared_ptr<CPUAllocator> m_cpuInputAllocator;
         std::shared_ptr<CPUAllocator> m_cpuOutputAllocator;
         std::shared_ptr<onnxruntime::KernelRegistry> m_kernelRegistry;
-        std::shared_ptr<const winrt::Windows::AI::MachineLearning::implementation::GraphNodeFactoryMap> m_graphNodeFactoryMap;
+        std::shared_ptr<const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfoMap> m_internalRegInfoMap;
         mutable uint64_t m_partitionKernelPrefixVal = 0;
 
         bool m_closed = false;
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.cpp
index 7a924e0c6ce81..8b6b42c63a70b 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.cpp
@@ -127,7 +127,7 @@ namespace Dml::GraphDescBuilder
             const onnxruntime::Node& node = *graph.GetNode(sortedNodeIndex);
 
             const GraphNodeProperties& graphNodeProps = graphNodePropertyMap.find(GetUniqueNodeName(node))->second;
-            const auto& requiredConstantCpuInputs = graphNodeProps.graphNodeFactoryRegistration->requiredConstantCpuInputs;
+            const auto& requiredConstantCpuInputs = graphNodeProps.internalRegInfo->requiredConstantCpuInputs;
 
             MLOperatorTensorGetter constantCpuNodeInputGetter = [&node, &constantCpuGraphInputGetter, &requiredConstantCpuInputs](uint32_t inputIndex)
             {
@@ -144,7 +144,7 @@ namespace Dml::GraphDescBuilder
             };
 
             DmlGraphNodeCreateInfo graphNodeInfo;
-            graphNodeProps.graphNodeFactoryRegistration->factory(
+            graphNodeProps.internalRegInfo->graphNodeFactoryRegistration->factory(
                 node,
                 constantCpuNodeInputGetter,
                 executionHandle,
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h
index 9d14a07545646..68fc7cc9f513d 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h
@@ -9,8 +9,8 @@ namespace Dml
 {
     struct GraphNodeProperties
     {
-        std::shared_ptr<const winrt::Windows::AI::MachineLearning::implementation::GraphNodeFactoryRegistration>
-        graphNodeFactoryRegistration;
+        std::shared_ptr<const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfo> 
+            internalRegInfo;
 
         // These are currently passed from the partitioning step since the only DML operators current 
         // supporting graph nodes don't customize the order of edges or shapes, other than coercing
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.cpp
index 6d1c16420aca9..12973981d75af 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.cpp
@@ -172,14 +172,14 @@ namespace Dml
         return true;
     }
 
-    bool NodeTensorTypesSupportedInGraph(const onnxruntime::Node& node, const GraphNodeFactoryRegistration& registration)
+    bool NodeTensorTypesSupportedInGraph(const onnxruntime::Node& node, const InternalRegistrationInfo& registration)
     {
         for (size_t i = 0; i < node.InputDefs().size(); ++i)
         {
             bool isConstantCpuInput = std::find(registration.requiredConstantCpuInputs.begin(), registration.requiredConstantCpuInputs.end(), i) !=
                   registration.requiredConstantCpuInputs.end();
 
-            if (!isConstantCpuInput && !NodeArgSupportedInGraph(node.InputDefs()[i], registration.requiresFloatFormatsExceptConstInputs))
+            if (!isConstantCpuInput && !NodeArgSupportedInGraph(node.InputDefs()[i], registration.graphNodeFactoryRegistration->requiresFloatFormatsExceptConstInputs))
             {
                 return false;
             }
@@ -187,7 +187,7 @@ namespace Dml
 
         for (auto arg : node.OutputDefs())
         {
-            if (!NodeArgSupportedInGraph(arg, registration.requiresFloatFormatsExceptConstInputs))
+            if (!NodeArgSupportedInGraph(arg, registration.graphNodeFactoryRegistration->requiresFloatFormatsExceptConstInputs))
             {
                 return false;
             }
@@ -196,8 +196,31 @@ namespace Dml
         return true;
     }
 
+    bool TryGetTensorDataType(
+        const onnxruntime::NodeArg& nodeArg,
+        _Out_ MLOperatorTensorDataType* onnxElementType
+    )
+    {
+        *onnxElementType = MLOperatorTensorDataType::Undefined;
+
+        const ::onnx::TypeProto* typeProto = nodeArg.TypeAsProto();
+        if (typeProto != nullptr && typeProto->has_tensor_type())
+        {
+            const ::onnx::TypeProto_Tensor& tensorTypeProto = typeProto->tensor_type();
+            if (tensorTypeProto.has_elem_type())
+            {
+                *onnxElementType = static_cast<MLOperatorTensorDataType>(tensorTypeProto.elem_type());
+                return true;
+            }
+        }
+
+        return false;
+    }
+
     bool DoesNodeContainSupportedDataTypes(
         const onnxruntime::Node& node,
+        const std::unordered_map<std::string, GraphPartition*>& nodeNameToPartitionMap,
+        _In_opt_ const InternalRegistrationInfo* regInfo,
         uint32_t supportedDeviceDataTypeMask // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         )
     {
@@ -209,33 +232,58 @@ namespace Dml
         {
             // Get the tensor element data type for this node, comparing against what the device actually supports.
             // Use the enumeration from the proto instead of nodeArg.Type() which returns a string.
-
-            const ::onnx::TypeProto* typeProto = nodeArg.TypeAsProto();
-            if (typeProto != nullptr && typeProto->has_tensor_type())
+            MLOperatorTensorDataType onnxElementType;
+            if (TryGetTensorDataType(nodeArg, &onnxElementType))
             {
-                const ::onnx::TypeProto_Tensor& tensorTypeProto = typeProto->tensor_type();
-                if (tensorTypeProto.has_elem_type())
+                DML_TENSOR_DATA_TYPE dmlElementType = GetDmlDataTypeFromMlDataTypeNoThrow(onnxElementType);
+                if (dmlElementType != DML_TENSOR_DATA_TYPE_UNKNOWN)
                 {
-                    MLOperatorTensorDataType onnxElementType = static_cast<MLOperatorTensorDataType>(tensorTypeProto.elem_type());
-                    DML_TENSOR_DATA_TYPE dmlElementType = GetDmlDataTypeFromMlDataTypeNoThrow(onnxElementType);
-                    if (dmlElementType != DML_TENSOR_DATA_TYPE_UNKNOWN)
+                    if (((1 << dmlElementType) & supportedDeviceDataTypeMask) == 0)
                     {
-                        if ((1 << dmlElementType) & supportedDeviceDataTypeMask)
-                        {
-                            // Leave nodeContainsSupportedDataTypes alone, since data type is supported.
-                            return;
-                        }
+                        nodeContainsSupportedDataTypes = false;
                     }
                 }
             }
-
-            // Else it's not supported (non-tensors, opaque data types, unsupported data types...).
-            nodeContainsSupportedDataTypes = false;
         };
 
         // Check whether the node uses any data types which are unsupported by the device.
         node.ForEachDef(nodeCallback);
 
+        // DML kernels support int64 and uint64 are expected to not *introduce* values out of range, which allows
+        // the temporary trick using strides to emulate 64 bit tensors to work.  If the source is a CPU operator, 
+        // graph input or initializer, it's not safe to assume the input can be represented with 32 bits.
+        if (regInfo)
+        {
+            for (uint32_t i = 0; i < node.InputDefs().size(); ++i) 
+            {
+                const auto* arg = node.InputDefs()[i];
+                MLOperatorTensorDataType onnxElementType;
+                if (arg->Exists() && TryGetTensorDataType(*arg, &onnxElementType))
+                {
+                    if (((onnxElementType == MLOperatorTensorDataType::UInt64) || (onnxElementType == MLOperatorTensorDataType::Int64)))
+                    {
+                        // Look up the input partition.  If it's a graph input or initializer it will be missing
+                        // from the map.  In this case or if the input comes from a CPU partition, it might be 
+                        // out of range.
+                        const std::string& argName = arg->Name();     
+                        auto partitionIter = nodeNameToPartitionMap.find(argName);
+                        if (partitionIter == nodeNameToPartitionMap.end() || !partitionIter->second->IsDmlPartition())
+                        {
+                            // Check if the operator handles the input on the CPU as a constant input
+                            bool isConstantCpuInput = std::find(regInfo->requiredConstantCpuInputs.begin(), regInfo->requiredConstantCpuInputs.end(), i) !=
+                                  regInfo->requiredConstantCpuInputs.end();
+
+                            if (!isConstantCpuInput)
+                            {
+                                nodeContainsSupportedDataTypes = false;
+                                break;
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
         return nodeContainsSupportedDataTypes;
     }
 
@@ -245,7 +293,8 @@ namespace Dml
         const onnxruntime::Node& node,    
         const std::vector<const onnxruntime::KernelRegistry*>& dmlRegistries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
-        const GraphNodeFactoryMap& graphNodeFactoryMap,
+        const InternalRegistrationInfoMap& internalRegInfoMap,
+        const std::unordered_map<std::string, GraphPartition*>& nodeNameToPartitionMap,
         _Inout_ std::unordered_map<const onnxruntime::Node*, GraphNodeProperties>& dmlNodePropertyMap,
         _Inout_ std::unordered_set<std::string>& requiredInitializerMap,
         _Out_ bool* isDmlNode,
@@ -260,11 +309,26 @@ namespace Dml
         for (auto registry : dmlRegistries) 
         {
             const onnxruntime::KernelCreateInfo* createInfo = registry->TryFindKernel(node, onnxruntime::kDmlExecutionProvider);
+            if (!createInfo)
+            {
+                continue;
+            }
+
+            auto regInfoIter = internalRegInfoMap.find(createInfo->kernel_def.get());
+            std::shared_ptr<InternalRegistrationInfo> internalRegInfo;
+            if (regInfoIter != internalRegInfoMap.end())
+            {
+                internalRegInfo = regInfoIter->second;
+                if (internalRegInfo->supportQuery && !internalRegInfo->supportQuery(node))
+                {
+                    continue;
+                }
+            }
 
             // Check whether the node uses any data types which are unsupported by the device.
-            bool nodeContainsSupportedDataTypes = DoesNodeContainSupportedDataTypes(node, supportedDeviceDataTypeMask);
+            bool nodeContainsSupportedDataTypes = DoesNodeContainSupportedDataTypes(node, nodeNameToPartitionMap, internalRegInfo.get(), supportedDeviceDataTypeMask);
 
-            if (createInfo && nodeContainsSupportedDataTypes)
+            if (nodeContainsSupportedDataTypes)
             {
                 *isDmlNode = true;
 
@@ -274,12 +338,11 @@ namespace Dml
 
                 // Ensure that shape information is known statically for the inputs and outputs of the node,
                 // which is required for MLGraph compilation.
-                auto graphNodeFactorMapIter = graphNodeFactoryMap.find(createInfo->kernel_def.get());
-                if (graphNodeFactorMapIter != graphNodeFactoryMap.end() && 
-                    NodeTensorTypesSupportedInGraph(node, *graphNodeFactorMapIter->second))
+                if (internalRegInfo && internalRegInfo->graphNodeFactoryRegistration && 
+                    NodeTensorTypesSupportedInGraph(node, *internalRegInfo))
                 {
                     bool requiredCpuInputsConstant = true;
-                    for (uint32_t inputIndex : graphNodeFactorMapIter->second->requiredConstantCpuInputs)
+                    for (uint32_t inputIndex : internalRegInfo->requiredConstantCpuInputs)
                     {
                         if (inputIndex >= node.InputDefs().size() || !node.InputDefs()[inputIndex]->Exists())
                         {
@@ -298,7 +361,7 @@ namespace Dml
                         requiredInitializerMap.insert(inputName);
                     }
 
-                    std::optional<uint32_t> requiredInputCount = graphNodeFactorMapIter->second->requiredInputCount;
+                    std::optional<uint32_t> requiredInputCount = internalRegInfo->graphNodeFactoryRegistration->requiredInputCount;
                     if (requiredCpuInputsConstant &&
                         TryGetStaticInputShapes( node, graphNodeProperty.first->second.inputShapes) &&
                         !ContainsEmptyDimensions(graphNodeProperty.first->second.inputShapes) &&
@@ -307,7 +370,7 @@ namespace Dml
                         (requiredInputCount == std::nullopt || *requiredInputCount == node.InputDefs().size()))
                     {
                         *isDmlGraphNode = true;
-                         graphNodeProperty.first->second.graphNodeFactoryRegistration = graphNodeFactorMapIter->second;
+                         graphNodeProperty.first->second.internalRegInfo = internalRegInfo;
                     }
                 }
 
@@ -550,7 +613,7 @@ namespace Dml
     std::vector<std::unique_ptr<GraphPartition>>
     BuildPartitions(
         const onnxruntime::GraphViewer& graph,
-        const GraphNodeFactoryMap& graphNodeFactoryMap,
+        const InternalRegistrationInfoMap& internalRegInfoMap,
         const std::vector<const onnxruntime::KernelRegistry*>& registries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         std::unordered_map<const onnxruntime::Node*, GraphNodeProperties>& graphNodePropertyMap,
@@ -610,7 +673,8 @@ namespace Dml
                 node,
                 registries,
                 supportedDeviceDataTypeMask,
-                graphNodeFactoryMap,
+                internalRegInfoMap,
+                nodeNameToPartitionMap,
                 graphNodePropertyMap,
                 requiredInitializerMap,
                 /*out*/ &isDmlNode,
@@ -726,7 +790,7 @@ namespace Dml
     std::vector<std::unique_ptr<onnxruntime::ComputeCapability>>
     PartitionGraph(
         const onnxruntime::GraphViewer& graph,
-        const GraphNodeFactoryMap& graphNodeFactoryMap,
+        const InternalRegistrationInfoMap& internalRegInfoMap,
         const std::vector<const onnxruntime::KernelRegistry*>& registries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         onnxruntime::KernelRegistry* registryForPartitionKernels,
@@ -741,7 +805,7 @@ namespace Dml
         std::unordered_map<const onnxruntime::Node*, GraphNodeProperties> graphNodePropertyMap;
         std::vector<std::unique_ptr<GraphPartition>> partitions = BuildPartitions(
             graph,
-            graphNodeFactoryMap, 
+            internalRegInfoMap, 
             registries,
             supportedDeviceDataTypeMask,
             graphNodePropertyMap, 
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h
index 4cc77ab426414..2227a2e67bf0a 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h
@@ -43,7 +43,7 @@ namespace Dml
     std::vector<std::unique_ptr<GraphPartition>>
     BuildPartitions(
         const onnxruntime::GraphViewer& graph,
-        const winrt::Windows::AI::MachineLearning::implementation::GraphNodeFactoryMap& graphNodeFactoryMap,
+        const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfoMap& internalRegInfoMap,
         const std::vector<const onnxruntime::KernelRegistry*>& registries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         std::unordered_map<const onnxruntime::Node*, GraphNodeProperties>& graphNodePropertyMap,
@@ -53,7 +53,7 @@ namespace Dml
     std::vector<std::unique_ptr<onnxruntime::ComputeCapability>>
     PartitionGraph(
         const onnxruntime::GraphViewer& graph,
-        const winrt::Windows::AI::MachineLearning::implementation::GraphNodeFactoryMap& graphNodeFactoryMap,
+        const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfoMap& internalRegInfoMap,
         const std::vector<const onnxruntime::KernelRegistry*>& registries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         onnxruntime::KernelRegistry* registryForPartitionKernels,
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.cpp
index 59317beb3cedb..b959e0c930755 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.cpp
@@ -1794,6 +1794,13 @@ HRESULT STDMETHODCALLTYPE MLKernelInferenceContext::SetOutputTensorShape(
 }
 CATCH_RETURN();
 
+MLSupportQueryContext::MLSupportQueryContext(
+        onnxruntime::OpNodeProtoHelper<onnxruntime::ProtoHelperNodeContext>* info,
+        const AttributeMap* defaultAttributes) : 
+    OpNodeInfoWrapper(info, nullptr, defaultAttributes, gsl::span<const uint32_t>(), MLOperatorTensorGetter())
+{
+}
+
 bool TryGetStaticShapeIfTensor(
     const onnx::TypeProto* inputProto,
     std::vector<uint32_t>& shapeDims) {
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.h
index c23eec22fe824..d6fabb2f7287b 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/MLOperatorAuthorImpl.h
@@ -616,6 +616,21 @@ void InferAndVerifyOutputSizes(
     const EdgeShapes* inputShapes,
     EdgeShapes& outputShapes);
 
+class MLSupportQueryContext final : public OpNodeInfoWrapper<
+    onnxruntime::ProtoHelperNodeContext,
+    WRL::Base<Microsoft::WRL::ChainInterfaces<IMLOperatorSupportQueryContextPrivate, IMLOperatorAttributes, IMLOperatorAttributes1>>,
+    onnxruntime::null_type>
+{
+ public:
+    MLSupportQueryContext() = delete;
+
+    MLSupportQueryContext(
+            onnxruntime::OpNodeProtoHelper<onnxruntime::ProtoHelperNodeContext>* info,
+            const AttributeMap* defaultAttributes);
+
+    // TODO - ...
+};
+
 onnxruntime::MLDataType ToTensorDataType(::MLOperatorTensorDataType type);
 std::string ToTypeString(MLOperatorEdgeDescription desc);
 onnx::AttributeProto_AttributeType ToProto(MLOperatorAttributeType type);
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorConvolution.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorConvolution.cpp
index 990e0981c8172..6a24b9c704301 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorConvolution.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorConvolution.cpp
@@ -14,14 +14,17 @@ class DmlOperatorConvolution : public DmlOperator, public ConvolutionHelperBase
     DmlOperatorConvolution(
         const MLOperatorKernelCreationContext& kernelInfo,
         DML_CONVOLUTION_MODE mode,
-        DML_CONVOLUTION_DIRECTION direction
+        DML_CONVOLUTION_DIRECTION direction,
+        bool hasDynamicPads
         )
     :   DmlOperator(kernelInfo),
-        ConvolutionHelperBase(kernelInfo, kernelInfo.GetTensorShapeDescription(), direction == DML_CONVOLUTION_DIRECTION_BACKWARD)
+        ConvolutionHelperBase(kernelInfo, kernelInfo.GetTensorShapeDescription(), direction == DML_CONVOLUTION_DIRECTION_BACKWARD, hasDynamicPads)
     {
-        ML_CHECK_VALID_ARGUMENT(kernelInfo.GetInputCount() >= 2);
+        uint32_t biasIndex = hasDynamicPads ? 3 : 2;
+        bool hasBiasInput = kernelInfo.GetInputCount() > biasIndex;
+
+        std::vector<std::optional<uint32_t>> kernelInputIndices = { 0, 1, biasIndex };
 
-        std::vector<std::optional<uint32_t>> kernelInputIndices = {0, 1, 2};
         DmlOperator::Initialize(kernelInfo, kernelInputIndices);
 
         // Vibranium DirectML is limited to handle only 2D and 3D convolution (4D and 5D tensors). So for 1D tensors,
@@ -32,7 +35,7 @@ class DmlOperatorConvolution : public DmlOperator, public ConvolutionHelperBase
         m_inputTensorDescs[1] = CreateTensorDescFromInput(kernelInfo, 1, TensorAxis::DoNotCoerce, TensorAxis::NoPlacementAdjustment, NonspatialDimensionCount, std::nullopt);
 
         // Bias is optional so only adjust it if it exists.
-        if (kernelInfo.GetInputCount() > 2)
+        if (hasBiasInput)
         {
             uint32_t inputDimSize = kernelInfo.GetTensorShapeDescription().GetInputTensorDimensionCount(0);
             ML_CHECK_VALID_ARGUMENT(
@@ -43,9 +46,9 @@ class DmlOperatorConvolution : public DmlOperator, public ConvolutionHelperBase
 
             // Resize the bias to be the same dimension as the input tensor.
             // The 1D tensor needs to be moved to the C channel.
-            m_inputTensorDescs[2] = CreateTensorDescFromInput(
+            m_inputTensorDescs[biasIndex] = CreateTensorDescFromInput(
                 kernelInfo, 
-                2, 
+                biasIndex, 
                 TensorAxis::DoNotCoerce, 
                 TensorAxis::C,
                 TensorAxis::LeftAligned,
@@ -73,7 +76,7 @@ class DmlOperatorConvolution : public DmlOperator, public ConvolutionHelperBase
         DML_CONVOLUTION_OPERATOR_DESC convDesc = {};
         convDesc.InputTensor = &inputDescs[0];
         convDesc.FilterTensor = &inputDescs[1];
-        convDesc.BiasTensor = kernelInfo.GetInputCount() > 2 ? &inputDescs[2] : nullptr;
+        convDesc.BiasTensor = hasBiasInput ? &inputDescs[biasIndex] : nullptr;
         convDesc.OutputTensor = &outputDescs[0];
         convDesc.Mode = mode;
         convDesc.Direction = direction;
@@ -92,19 +95,20 @@ class DmlOperatorConvolution : public DmlOperator, public ConvolutionHelperBase
 };
 
 // A specific type of operation for registration.
-template <DML_CONVOLUTION_MODE Mode, DML_CONVOLUTION_DIRECTION Direction>
+template <DML_CONVOLUTION_MODE Mode, DML_CONVOLUTION_DIRECTION Direction, bool hasDynamicPads = false>
 class DmlOperatorConvolutionTemplate : public DmlOperatorConvolution
 {
 public:
     DmlOperatorConvolutionTemplate(const MLOperatorKernelCreationContext& kernelInfo)
-    :   DmlOperatorConvolution(kernelInfo, Mode, Direction)
+    :   DmlOperatorConvolution(kernelInfo, Mode, Direction, hasDynamicPads)
     {
     }
 };
 
-DML_OP_DEFINE_CREATION_FUNCTION(Conv,               DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_FORWARD>);
-DML_OP_DEFINE_CREATION_FUNCTION(ConvTranspose,      DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_BACKWARD>);
-DML_OP_DEFINE_CREATION_FUNCTION(FusedConv,          DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_FORWARD>);
-DML_OP_DEFINE_CREATION_FUNCTION(FusedConvTranspose, DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_BACKWARD>);
+DML_OP_DEFINE_CREATION_FUNCTION(Conv,                           DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_FORWARD>);
+DML_OP_DEFINE_CREATION_FUNCTION(ConvTranspose,                  DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_BACKWARD>);
+DML_OP_DEFINE_CREATION_FUNCTION(FusedConv,                      DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_FORWARD>);
+DML_OP_DEFINE_CREATION_FUNCTION(FusedConvTranspose,             DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_BACKWARD>);
+DML_OP_DEFINE_CREATION_FUNCTION(ConvTransposeWithDynamicPads,   DmlOperatorConvolutionTemplate<DML_CONVOLUTION_MODE_CROSS_CORRELATION, DML_CONVOLUTION_DIRECTION_BACKWARD, true>);
 
 } // namespace Dml
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorCopy.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorCopy.cpp
index 0aeb4a2fc0959..eb985f6c404c9 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorCopy.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorCopy.cpp
@@ -24,10 +24,7 @@ class DmlOperatorCopy : public DmlOperator
         // element counts are the same. All this operator does is copy the resource and
         // rearrange the dimensions, so we tell DML that the output dimensions are the
         // same as the input dimensions.
-        m_outputTensorDescs.front() = TensorDesc(
-            m_outputTensorDescs.front().GetDmlDataType(), 
-            m_inputTensorDescs.front().GetSizes()
-            );
+        m_outputTensorDescs.front() = m_inputTensorDescs.front();
 
         ComPtr<IMLOperatorKernelCreationContextPrivate> contextPrivate;
         THROW_IF_FAILED(kernelInfo.GetInterface()->QueryInterface(contextPrivate.GetAddressOf()));
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorPooling.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorPooling.cpp
index 9bd86a8d43b94..7f27291ede562 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorPooling.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorPooling.cpp
@@ -29,6 +29,19 @@ class DmlOperatorPooling : public DmlOperator, public PoolingHelperBase
 
         assert(m_kernel.spatialDimensionCount <= ARRAYSIZE(m_kernel.windowSize));
 
+        // The below attributes are temporarily not supported:
+        int ceilMode = kernelInfo.GetOptionalAttribute<int>(AttrName::CeilMode, 0);
+        THROW_HR_IF(E_NOTIMPL, ceilMode != 0);
+
+        int storageOrder = kernelInfo.GetOptionalAttribute<int>(AttrName::StorageOrder, 0);
+        THROW_HR_IF(E_NOTIMPL, storageOrder != 0);
+
+        auto dilations = kernelInfo.GetOptionalAttributeVectorInt32(AttrName::Dilations);
+        for (int dilation : dilations)
+        {
+            THROW_HR_IF(E_NOTIMPL, dilation != 1);
+        }
+
         // DML requires that DimensionCount be equal to Input.DimCount - 2 for Pooling
         uint32_t expectedSpatialDimCount = m_inputTensorDescs[0].GetDimensionCount() - 2;
         if (m_kernel.spatialDimensionCount < expectedSpatialDimCount)
@@ -121,6 +134,37 @@ class DmlOperatorPoolingTemplate : public DmlOperatorPooling
     }
 };
 
+void QueryMaxPool(IMLOperatorSupportQueryContextPrivate* context, bool *isSupported)
+{
+    *isSupported = false;
+    
+    MLOperatorAttributes attributes(context);
+
+    // The below attributes are temporarily not supported:
+    int ceilMode = attributes.GetOptionalAttribute<int>(AttrName::CeilMode, 0);
+    if (ceilMode != 0)
+    {
+        return;
+    }
+
+    int storageOrder = attributes.GetOptionalAttribute<int>(AttrName::StorageOrder, 0);
+    if (storageOrder != 0)
+    {
+        return;
+    }
+
+    auto dilations = attributes.GetOptionalAttributeVectorInt32(AttrName::Dilations);
+    for (int dilation : dilations)
+    {
+        if (dilation != 1)
+        {
+            return;
+        }
+    }
+
+    *isSupported = true;
+}
+
 DML_OP_DEFINE_CREATION_FUNCTION(AveragePool,           DmlOperatorPoolingTemplate<DML_OPERATOR_AVERAGE_POOLING, false>);
 DML_OP_DEFINE_CREATION_FUNCTION(GlobalAveragePool,     DmlOperatorPoolingTemplate<DML_OPERATOR_AVERAGE_POOLING, true>);
 DML_OP_DEFINE_CREATION_FUNCTION(MaxPool,               DmlOperatorPoolingTemplate<DML_OPERATOR_MAX_POOLING1, false>);
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorSlice.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorSlice.cpp
index 7ae1995e74cbe..e167a89f0606e 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorSlice.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorSlice.cpp
@@ -6,16 +6,22 @@
 namespace Dml
 {
 
-class DmlOperatorSlice : public DmlOperator, public SliceHelper
+class DmlOperatorSlice : public DmlOperator, public SliceHelperBase
 {
 public:
-    DmlOperatorSlice(const MLOperatorKernelCreationContext& kernelInfo)
+    DmlOperatorSlice(const MLOperatorKernelCreationContext& kernelInfo, uint32_t opsetVersion)
     :   DmlOperator(kernelInfo),
-        SliceHelper(kernelInfo, kernelInfo.GetTensorShapeDescription())
+        SliceHelperBase(kernelInfo, kernelInfo.GetTensorShapeDescription(), opsetVersion)
     {
-        ML_CHECK_VALID_ARGUMENT(kernelInfo.GetInputCount() == 1);
+        uint32_t minInputCount = (opsetVersion < 10) ? 1 : 3;
+        ML_CHECK_VALID_ARGUMENT(kernelInfo.GetInputCount() >= minInputCount);
         ML_CHECK_VALID_ARGUMENT(kernelInfo.GetOutputCount() == 1);
-        DmlOperator::Initialize(kernelInfo);
+
+        // TODO (23108599): Slice V10 introduces an optional "Steps" input which the kernel does not yet support.
+        THROW_HR_IF(E_NOTIMPL, kernelInfo.GetInputCount() > 4);
+
+        std::vector<std::optional<uint32_t>> kernelInputIndices = { 0 };
+        DmlOperator::Initialize(kernelInfo, kernelInputIndices);
 
         assert(m_inputTensorDescs[0].GetDimensionCount() >= gsl::narrow_cast<uint32_t>(m_offsets.size()));
         assert(m_inputTensorDescs[0].GetDimensionCount() >= gsl::narrow_cast<uint32_t>(m_sizes.size()));
@@ -54,6 +60,22 @@ class DmlOperatorSlice : public DmlOperator, public SliceHelper
     }
 };
 
-DML_OP_DEFINE_CREATION_FUNCTION(Slice, DmlOperatorSlice);
+// A specific type of operation for registration.
+template <uint32_t opsetVersion>
+class DmlOperatorSliceTemplate : public DmlOperatorSlice
+{
+public:
+    DmlOperatorSliceTemplate(const MLOperatorKernelCreationContext& kernelInfo)
+    :   DmlOperatorSlice(kernelInfo, opsetVersion)
+    {
+    }
+};
+
+void QuerySlice(IMLOperatorSupportQueryContextPrivate* context, bool *isSupported)
+{
+    *isSupported = (context->GetInputCount() <= 4);
+}
 
+DML_OP_DEFINE_CREATION_FUNCTION(Slice7,  DmlOperatorSliceTemplate<7>);
+DML_OP_DEFINE_CREATION_FUNCTION(Slice10, DmlOperatorSliceTemplate<10>);
 } // namespace Dml
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.cpp b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.cpp
index b35c570bf8c53..c0ade42ccd5a4 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.cpp
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.cpp
@@ -70,12 +70,14 @@ struct OperatorRegistrationInformation
     // can't be represented as nodes in an optimized graph yet.
     std::optional<uint32_t> requiredInputCountForDmlGraphSupport;
 
+    MLOperatorSupportQueryFunction supportQueryFunction;
 };
 
 DML_OP_EXTERN_CREATION_FUNCTION(Copy);
 DML_OP_EXTERN_CREATION_FUNCTION(FC);
 DML_OP_EXTERN_CREATION_FUNCTION(Conv);
 DML_OP_EXTERN_CREATION_FUNCTION(ConvTranspose);
+DML_OP_EXTERN_CREATION_FUNCTION(ConvTransposeWithDynamicPads);
 DML_OP_EXTERN_CREATION_FUNCTION(AveragePool);
 DML_OP_EXTERN_CREATION_FUNCTION(GlobalAveragePool);
 DML_OP_EXTERN_CREATION_FUNCTION(MaxPool);
@@ -97,7 +99,8 @@ DML_OP_EXTERN_CREATION_FUNCTION(Split);
 DML_OP_EXTERN_CREATION_FUNCTION(Transpose);
 DML_OP_EXTERN_CREATION_FUNCTION(Tile);
 DML_OP_EXTERN_CREATION_FUNCTION(Concat);
-DML_OP_EXTERN_CREATION_FUNCTION(Slice);
+DML_OP_EXTERN_CREATION_FUNCTION(Slice7);
+DML_OP_EXTERN_CREATION_FUNCTION(Slice10);
 DML_OP_EXTERN_CREATION_FUNCTION(Pad);
 DML_OP_EXTERN_CREATION_FUNCTION(SpaceToDepth);
 DML_OP_EXTERN_CREATION_FUNCTION(DepthToSpace);
@@ -201,6 +204,9 @@ DML_OP_EXTERN_CREATION_FUNCTION(Scatter);
 DML_OP_EXTERN_CREATION_FUNCTION(Resize);
 DML_OP_EXTERN_CREATION_FUNCTION(ConstantOfShape);
 
+DML_OP_EXTERN_QUERY_FUNCTION(MaxPool);
+DML_OP_EXTERN_QUERY_FUNCTION(Slice);
+
 const static char* const typeNameListDefault[1] = {"T"};
 const static char* const typeNameListTopK[2] = { "T", "I" };
 const static char* const typeNameListLogicalComparison[2] = { "T", "T1" };
@@ -220,7 +226,7 @@ const static SupportedTensorDataTypes supportedTypeListAllScalars[1] = { Support
 const static SupportedTensorDataTypes supportedTypeListBool[1] = {SupportedTensorDataTypes::Bool};
 const static SupportedTensorDataTypes supportedTypeListTopK[2] = {SupportedTensorDataTypes::Float16to32, SupportedTensorDataTypes::Int64};
 const static SupportedTensorDataTypes supportedTypeListIndices[1] = { SupportedTensorDataTypes::Int32|SupportedTensorDataTypes::Int64 };
-const static SupportedTensorDataTypes supportedTypeListCast[2] = { SupportedTensorDataTypes::Scalars8to32, SupportedTensorDataTypes::Scalars8to32 };
+const static SupportedTensorDataTypes supportedTypeListCast[2] = { SupportedTensorDataTypes::AllScalars, SupportedTensorDataTypes::Scalars8to32 };
 const static SupportedTensorDataTypes supportedTypeListScatterGather[2] = { SupportedTensorDataTypes::Float16to32, SupportedTensorDataTypes::Int32 | SupportedTensorDataTypes::Int64 };
 const static SupportedTensorDataTypes supportedTypeListQuantize[2] = { SupportedTensorDataTypes::Float32, SupportedTensorDataTypes::UInt8 };
 const static SupportedTensorDataTypes supportedTypeListIsNan[2] = { SupportedTensorDataTypes::Float16to32, SupportedTensorDataTypes::UInt8 };
@@ -233,6 +239,10 @@ const static SupportedTensorDataTypes supportedTypeListLogicalComparison[2] = /*
 #define REG_INFO(version, operatorName, ...) \
     #operatorName, OnnxOperatorSet##version::sc_sinceVer_##operatorName, onnxruntime::kOnnxDomain, Create##operatorName, ShapeInferenceFunction<ShapeInferenceHelper_##operatorName>, false, false, ##__VA_ARGS__, 
 
+// Versioned operator
+#define REG_INFO_VER(version, operatorName, ...) \
+    #operatorName, OnnxOperatorSet##version::sc_sinceVer_##operatorName, onnxruntime::kOnnxDomain, Create##operatorName##version, ShapeInferenceFunction<ShapeInferenceHelper_##operatorName##version>, false, false, ##__VA_ARGS__, 
+
 // Identity operators use Copy, alias their first input, and require floating point formats
 // for usage in the graph, besides constant inputs.  This is because they currently use 
 // element-wise identity operators  in the graph for striding support, but issue actual copies 
@@ -242,12 +252,17 @@ const static SupportedTensorDataTypes supportedTypeListLogicalComparison[2] = /*
 
 // MS-domain operators
 #define REG_INFO_MS(version, operatorName, ...) \
+    #operatorName, MsftOperatorSet##version::sc_sinceVer_##operatorName, onnxruntime::kMSDomain, Create##operatorName, ShapeInferenceFunction<ShapeInferenceHelper_##operatorName>, false, false, ##__VA_ARGS__, 
+
+// MS-domain operators
+#define REG_INFO_MSDML(version, operatorName, ...) \
     #operatorName, MsftOperatorSet##version::sc_sinceVer_##operatorName, onnxruntime::kMSDmlDomain, Create##operatorName, ShapeInferenceFunction<ShapeInferenceHelper_##operatorName>, false, false, ##__VA_ARGS__, 
 
 const static OperatorRegistrationInformation operatorRegistrationInformationTable[] =
 {
 ///  Domain/Type, Ver,  Name,                               TypeNames,                       Types,                              Graph Support,                  Required const CPU inputs, 
-///                                                                                                                                                              Input count required for graph support
+///                                                                                                                                                              Input count required for graph support,
+///                                                                                                                                                              Support query function
 
     // Deep Learning Standard Layers
     {REG_INFO(      7,  Conv,                               typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
@@ -255,7 +270,9 @@ const static OperatorRegistrationInformation operatorRegistrationInformationTabl
     {REG_INFO(      7,  AveragePool,                        typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  GlobalAveragePool,                  typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  MaxPool,                            typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO(      8,  MaxPool,                            typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO(      8,  MaxPool,                            typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported, {}, std::nullopt, QueryMaxPool)},
+    {REG_INFO(      10, MaxPool,                            typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported, {}, std::nullopt, QueryMaxPool)},
+      
     {REG_INFO(      7,  GlobalMaxPool,                      typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  LpPool,                             typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  GlobalLpPool,                       typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
@@ -269,12 +286,14 @@ const static OperatorRegistrationInformation operatorRegistrationInformationTabl
     {REG_INFO(      7,  RNN,                                typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::NotSupported)},
     {REG_INFO(      7,  GRU,                                typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::NotSupported)},
     {REG_INFO(      7,  LSTM,                               typeNameListDefault,             supportedTypeListFloat16to32,       DmGraphSupport::NotSupported)},
+    {REG_INFO_MS(   1,  ConvTransposeWithDynamicPads,       typeNameListDefault,            supportedTypeListFloat16to32,        DmGraphSupport::Supported,      {2})},
 
     // Data Reorganization Layers
     {REG_INFO(      7,  Split,                              typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  Transpose,                          typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  Concat,                             typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO(      7,  Slice,                              typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_VER(  7,  Slice,                              typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_VER(  10, Slice,                              typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported,      {1, 2, 3}, std::nullopt, QuerySlice)},
     {REG_INFO(      7,  Pad,                                typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  SpaceToDepth,                       typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
     {REG_INFO(      7,  DepthToSpace,                       typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
@@ -400,19 +419,19 @@ const static OperatorRegistrationInformation operatorRegistrationInformationTabl
     {REG_INFO(      9,  OneHot,                             typeNameListOneHot,             supportedTypeListOneHot,            DmGraphSupport::Supported,      {1})},
                                                         
     // Fused operators                                  
-    {REG_INFO_MS(   1,  FusedConv,                          typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedConvTranspose,                 typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedInstanceNormalization,         typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedBatchNormalization,            typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedMeanVarianceNormalization,     typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedGemm,                          typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedMatMul,                        typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
-    {REG_INFO_MS(   1,  FusedAdd,                           typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)}, 
-    {REG_INFO_MS(   1,  FusedSum,                           typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported,      {}, 2)}, 
+    {REG_INFO_MSDML(1,  FusedConv,                          typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedConvTranspose,                 typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedInstanceNormalization,         typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedBatchNormalization,            typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedMeanVarianceNormalization,     typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedGemm,                          typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedMatMul,                        typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)},
+    {REG_INFO_MSDML(1,  FusedAdd,                           typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported)}, 
+    {REG_INFO_MSDML(1,  FusedSum,                           typeNameListDefault,            supportedTypeListFloat16to32,       DmGraphSupport::Supported,      {}, 2)}, 
                 
     // TODO: DwayneR implement MaxUnpool https://dev.azure.com/microsoft/OS/_workitems/edit/21267466
 };
-
+ 
 template<typename T> 
 MLOperatorEdgeDescription EdgeDesc()
 {
@@ -497,10 +516,17 @@ void RegisterDmlOperators(IMLOperatorRegistry* registry)
             shapeInferrer = wil::MakeOrThrow<MLOperatorShapeInferrer>(information.shapeInferenceFunction);
         }
 
+        ComPtr<IMLOperatorSupportQueryPrivate> supportQuery;
+        if (information.supportQueryFunction)
+        {
+            supportQuery = wil::MakeOrThrow<MLOperatorSupportQuery>(information.supportQueryFunction);
+        }
+
         THROW_IF_FAILED(registryPrivate->RegisterOperatorKernel(
             &desc, 
             factory.Get(), 
             shapeInferrer.Get(),
+            supportQuery.Get(),
             true, // isInternalOperator
             information.canAliasFirstInput, // alias
             kernelSupportsGraph, // supportsGraph
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.h
index 12ca021b60fc2..186608d78b586 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/OperatorRegistration.h
@@ -9,6 +9,7 @@ class MLOperatorKernelCreationContext;
 
 // Forward declares an external creation function.
 #define DML_OP_EXTERN_CREATION_FUNCTION(operatorName) extern void Create##operatorName(IMLOperatorKernelCreationContext* kernelInfo, IMLOperatorKernel** opKernel)
+#define DML_OP_EXTERN_QUERY_FUNCTION(operatorName) extern void Query##operatorName(IMLOperatorSupportQueryContextPrivate* context, bool *isSupported);
 
 // Declares a callback creation function of the given operator class.
 // This does not register it, just declares it for usage by registration later.
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/Attributes.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/Attributes.h
index 20f4d69cbdde0..dc7def086dad2 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/Attributes.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/Attributes.h
@@ -19,6 +19,7 @@ namespace AttrName
     static constexpr const char* BlockSize = "blocksize";
     static constexpr const char* Border = "border";
     static constexpr const char* Broadcast = "broadcast";
+    static constexpr const char* CeilMode = "ceil_mode";
     static constexpr const char* Clip = "clip";
     static constexpr const char* CountIncludePad = "count_include_pad";
     static constexpr const char* Dilations = "dilations";
@@ -60,6 +61,7 @@ namespace AttrName
     static constexpr const char* Split = "split";
     static constexpr const char* Starts = "starts";
     static constexpr const char* Steepness = "steepness";
+    static constexpr const char* StorageOrder = "storage_order";
     static constexpr const char* Strides = "strides";
     static constexpr const char* Tiles = "tiles";
     static constexpr const char* To = "to";
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorHelper.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorHelper.h
index 871df4070987f..b15955e0e533d 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorHelper.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorHelper.h
@@ -735,6 +735,7 @@ class MLOperatorKernel : public Microsoft::WRL::RuntimeClass<
 using MLOperatorTypeInferenceFunction = void (CALLBACK*)(IMLOperatorTypeInferenceContext*);
 using MLOperatorShapeInferenceFunction = void (CALLBACK*)(IMLOperatorShapeInferenceContext*);
 using MLOperatorKernelCreateFn = void(*)(IMLOperatorKernelCreationContext*, IMLOperatorKernel**);
+using MLOperatorSupportQueryFunction = void (CALLBACK*)(IMLOperatorSupportQueryContextPrivate*, bool*);
 
 class MLOperatorShapeInferrer : public Microsoft::WRL::RuntimeClass<
     Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>, IMLOperatorShapeInferrer>
@@ -755,6 +756,29 @@ class MLOperatorShapeInferrer : public Microsoft::WRL::RuntimeClass<
     MLOperatorShapeInferenceFunction m_shapeInferenceFn = nullptr;
 };
 
+class MLOperatorSupportQuery : public Microsoft::WRL::RuntimeClass<
+    Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>, IMLOperatorSupportQueryPrivate>
+{
+public:
+    MLOperatorSupportQuery(MLOperatorSupportQueryFunction queryFn) :
+        m_queryFn(queryFn)
+    {}
+
+    HRESULT STDMETHODCALLTYPE QuerySupport(
+        IMLOperatorSupportQueryContextPrivate* context,
+        BOOL* isSupported) noexcept override try
+    {
+        bool fIsSupported = false;
+        m_queryFn(context, &fIsSupported);
+        *isSupported = fIsSupported ? TRUE : FALSE;
+        return S_OK;
+    }
+    CATCH_RETURN();
+
+private:
+    MLOperatorSupportQueryFunction m_queryFn = nullptr;
+};
+
 class MLOperatorTypeInferrer : public Microsoft::WRL::RuntimeClass<
     Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>, IMLOperatorTypeInferrer>
 {
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorPrivate.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorPrivate.h
index 6b0cbab3fb1fd..216dfdcb2ba3d 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorPrivate.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/MLOperatorAuthorPrivate.h
@@ -46,6 +46,60 @@ IMLOperatorKernelCreationContextPrivate : public IMLOperatorKernelCreationContex
     ) const noexcept PURE;
 };
 
+//! \interface IMLOperatorAttributes1
+//! \brief Represents the values of an operator's attributes, as determined by a model using the operator.
+//! This interface is called by implementations of custom operator kernels, and by implementations
+//! of shape and type inferrers.
+interface DECLSPEC_UUID("3a798815-dfe3-4bcd-b6a6-f70650d5f80b") DECLSPEC_NOVTABLE
+IMLOperatorAttributes1 : public IMLOperatorAttributes
+{
+    //! Gets an interface pointer for the constant tensor.
+    //! Note the tensor is CPU side (IsCpuData is true).
+    STDMETHOD(GetTensorAttribute)(
+        _In_z_ const char* name,
+        _COM_Outptr_ IMLOperatorTensor** tensor
+        ) const noexcept PURE;
+};
+
+interface __declspec(uuid("897bb586-6cee-4106-8513-dda33151c109")) DECLSPEC_NOVTABLE
+IMLOperatorSupportQueryContextPrivate : public IMLOperatorAttributes1
+{
+    //! Gets the number of inputs to the operator.
+    STDMETHOD_(uint32_t, GetInputCount)() const noexcept PURE;
+
+    //! Gets the number of outputs to the operator.
+    STDMETHOD_(uint32_t, GetOutputCount)() const noexcept PURE;
+
+    //! Returns true if an input to the operator is valid.  
+    //! This always returns true except for optional inputs and invalid indices.
+    STDMETHOD_(bool, IsInputValid)(uint32_t inputIndex) const noexcept PURE;
+
+    //! Returns true if an output to the operator is valid.
+    //! This always returns true if within GetOutputCount except for optional outputs.
+    STDMETHOD_(bool, IsOutputValid)(uint32_t outputIndex) const noexcept PURE;
+
+    //! Gets the description of the specified input edge of the operator.
+    STDMETHOD(GetInputEdgeDescription)(
+        uint32_t inputIndex, 
+        _Out_ MLOperatorEdgeDescription* edgeDescription
+        ) const noexcept PURE;
+
+    //! Gets the description of the specified output edge of the operator.
+    STDMETHOD(GetOutputEdgeDescription)(
+        uint32_t outputIndex, 
+        _Out_ MLOperatorEdgeDescription* edgeDescription
+        ) const noexcept PURE;
+};
+
+interface __declspec(uuid("023954b3-aed2-4b03-b7c7-f0838053a9a1")) DECLSPEC_NOVTABLE
+IMLOperatorSupportQueryPrivate : public IUnknown
+{
+    STDMETHOD(QuerySupport)(
+        IMLOperatorSupportQueryContextPrivate* context,
+        BOOL* isSupported
+        ) noexcept PURE;
+};
+
 interface DECLSPEC_UUID("3de1dc1e-13e9-4099-ae88-7b4100083415") DECLSPEC_NOVTABLE
 IMLOperatorRegistryPrivate : public IUnknown
 {
@@ -53,6 +107,7 @@ IMLOperatorRegistryPrivate : public IUnknown
         const MLOperatorKernelDescription* operatorKernel,
         IMLOperatorKernelFactory* operatorKernelFactory,
         _In_opt_ IMLOperatorShapeInferrer* shapeInferrer,
+        _In_opt_ IMLOperatorSupportQueryPrivate* supportQuery,
         bool isInternalOperator,
         bool canAliasFirstInput,
         bool supportsGraph,
@@ -63,21 +118,6 @@ IMLOperatorRegistryPrivate : public IUnknown
         ) const noexcept PURE;
 };
 
-//! \interface IMLOperatorAttributes1
-//! \brief Represents the values of an operator's attributes, as determined by a model using the operator.
-//! This interface is called by implementations of custom operator kernels, and by implementations
-//! of shape and type inferrers.
-interface DECLSPEC_UUID("3a798815-dfe3-4bcd-b6a6-f70650d5f80b") DECLSPEC_NOVTABLE
-IMLOperatorAttributes1 : public IMLOperatorAttributes
-{
-    //! Gets an interface pointer for the constant tensor.
-    //! Note the tensor is CPU side (IsCpuData is true).
-    STDMETHOD(GetTensorAttribute)(
-        _In_z_ const char* name,
-        _COM_Outptr_ IMLOperatorTensor** tensor
-        ) const noexcept PURE;
-};
-
 // Declare private enum MLOperatorAttributeType::Tensor.
 //
 //      enum class MLOperatorAttributeType : uint32_t
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.cpp b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.cpp
index 99425a76f49f0..9f3a0ed7e69f4 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.cpp
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.cpp
@@ -430,58 +430,7 @@ int64_t ReadAsInt64(MLOperatorTensorDataType tensorDataType, const void* p)
         return edgeShapes;
     }
 
-    void SliceHelper::Initialize(
-        const MLOperatorAttributes& operatorAttributes,
-        gsl::span<const DimensionType> inputDimensions
-        )
-    {
-        const uint32_t dimCount = gsl::narrow_cast<int32_t>(inputDimensions.size());
-
-        std::vector<int32_t> starts = operatorAttributes.GetOptionalAttributeVectorInt32(AttrName::Starts);
-        std::vector<int32_t> ends = operatorAttributes.GetOptionalAttributeVectorInt32(AttrName::Ends);
-        std::vector<int32_t> axes = operatorAttributes.GetOptionalAttributeVectorInt32(AttrName::Axes);
-        HandleNegativeAxes(/*inout*/ axes, dimCount);
-
-        ML_CHECK_VALID_ARGUMENT(starts.size() == ends.size(), "'starts' must equal 'ends' in size.");
-        ML_CHECK_VALID_ARGUMENT(axes.empty() || starts.size() == axes.size(), "'axes' must equal 'starts' in size, or 'axes' must be empty.");
-
-        m_outputDimensions.assign(inputDimensions.begin(), inputDimensions.end());
-        m_offsets.resize(m_outputDimensions.size());
-        m_sizes.resize(m_outputDimensions.size());
-        m_strides.resize(m_outputDimensions.size(), 1); // Only a stride of 1 element is supported by ONNX 1.2.
-
-        // Set initial defaults lest 'starts' and 'ends' arrays are shorter than the dimension count.
-        std::copy(inputDimensions.begin(), inputDimensions.begin() + m_outputDimensions.size(), m_sizes.begin());
-
-        // Clamp selected dimensions to given 'starts' and 'ends'.
-        for (int i = 0, ci = gsl::narrow_cast<int>(starts.size()); i < ci; ++i)
-        {
-            int dimIndex = i;
-            if (!axes.empty())
-            {
-                dimIndex = axes[i];
-            }
-            ML_CHECK_VALID_ARGUMENT(dimIndex < inputDimensions.size(), "'axes' must be valid with within actual input dimensions.");
-
-            // Positive values are offsets from 0.
-            // Negative values are offsets from the dimension's size.
-            int dim   = gsl::narrow_cast<int>(inputDimensions[dimIndex]);
-            int start = (starts[i] < 0) ? (starts[i] + dim) : starts[i];
-            int end   = (ends[i]   < 0) ? (ends[i]   + dim) : ends[i];
-
-            // Clamp the dimensions to the slice extents.
-            // Clamp negative numbers to 0, per case test_slice_start_out_of_bounds.
-            start = std::max(start, 0);
-            end = std::min(end, dim);
-            int size = std::max(end - start, 0);
-
-            m_outputDimensions[dimIndex] = size;
-            m_offsets[dimIndex] = start;
-            m_sizes[dimIndex] = gsl::narrow_cast<uint32_t>(size);
-        }
-    }
-
-    std::vector<EdgeShapes> SliceHelper::GetOutputShapes(const MLShapeInferenceContext& shapeInfo) const
+    std::vector<EdgeShapes> SliceHelperBase::GetOutputShapes(const MLShapeInferenceContext& shapeInfo) const    
     {
         return { m_outputDimensions };
     }
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h
index ad6da14f76469..03c003d2b6d0d 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h
@@ -299,13 +299,13 @@ class ConvolutionHelperBase
 {
 public:
     enum FilterDims { K };
-    enum InputTensor { X, Filter, Bias };
+    enum InputTensor { X, Filter};
     enum InputDims { N, C, H, W };
 
 public:
     // Info_t is used to obtain attributes which will be used for calculating the output shape later. 
     template<typename Info_t, typename Shape_t>
-    ConvolutionHelperBase(const Info_t& info, const Shape_t& shape, bool transpose) :
+    ConvolutionHelperBase(const Info_t& info, const Shape_t& shape, bool transpose, bool hasDynamicPads) :
         m_kernel(InitializeKernel(info, shape.GetInputTensorDimensionCount(0), shape.GetInputTensorShape(1)))
     {
         m_groupCount = info.GetOptionalAttribute<uint32_t>(AttrName::Group, 1);
@@ -316,7 +316,7 @@ class ConvolutionHelperBase
         }
         else
         {
-            InitializeKernelAndShapesTransposed(info, shape);
+            InitializeKernelAndShapesTransposed(info, shape, hasDynamicPads);
         }
     }
 
@@ -347,7 +347,7 @@ class ConvolutionHelperBase
     
     
     template<typename Info_t, typename Shape_t>
-    void InitializeKernelAndShapesTransposed(const Info_t& info, const Shape_t& shapeInfo)
+    void InitializeKernelAndShapesTransposed(const Info_t& info, const Shape_t& shapeInfo, bool hasDynamicPads)
     {
         std::vector<int> outputShape = info.GetOptionalAttributeVectorInt32(AttrName::OutputShape);
         if (!outputShape.empty())
@@ -363,7 +363,33 @@ class ConvolutionHelperBase
 
         ML_CHECK_VALID_ARGUMENT(inputDimensions.size() > NonspatialDimensionCount, "Input dimensions must be >= 3");
 
-        ResolvingPadding(inputDimensions);
+        if (hasDynamicPads)
+        {
+            MLOperatorTensor padsTensor = info.GetConstantInputTensor(2);
+            const std::vector<uint32_t>& padsTensorDimensions = padsTensor.GetShape();
+            ML_CHECK_VALID_ARGUMENT(padsTensorDimensions.size() == 1, "Pads dimensions must equal 1");
+            const size_t dimCount = padsTensorDimensions[0];
+            ML_CHECK_VALID_ARGUMENT(dimCount == 2 * NchwSpatialDimensionCount, "Pads count must equal 4");
+            const int64_t* padsData = padsTensor.GetData<int64_t>();
+
+            for (size_t i = 0; i < dimCount; ++i)
+            {
+                ML_CHECK_VALID_ARGUMENT(padsData[i] >= 0, "Padding values must be greater than or equal to 0");
+                if (i < dimCount / 2)
+                {
+                    m_kernel.startPadding[i] = gsl::narrow_cast<uint32_t>(padsData[i]);
+                }
+                else
+                {
+                    m_kernel.endPadding[i - dimCount/2] = gsl::narrow_cast<uint32_t>(padsData[i]);
+                }
+            }
+        }
+        else
+        {
+            ResolvingPadding(inputDimensions);
+        }
+
         m_outputShapes.resize(1);
         m_outputShapes[0] = InitializeKernelOutputDimsTranspose(inputDimensions, m_kernel);
         static_assert(C < NonspatialDimensionCount);
@@ -419,14 +445,21 @@ class ConvHelper : public ConvolutionHelperBase
 {
 public:
     template<typename Info_t, typename Shape_t>
-    ConvHelper(const Info_t& info, const Shape_t& shape) : ConvolutionHelperBase(info, shape, false) {}
+    ConvHelper(const Info_t& info, const Shape_t& shape) : ConvolutionHelperBase(info, shape, false, false) {}
 };
 
 class ConvTransposeHelper : public ConvolutionHelperBase
 {
 public:
     template<typename Info_t, typename Shape_t>
-    ConvTransposeHelper(const Info_t& info, const Shape_t& shape) : ConvolutionHelperBase(info, shape, true) {}
+    ConvTransposeHelper(const Info_t& info, const Shape_t& shape) : ConvolutionHelperBase(info, shape, true, false) {}
+};
+
+class ConvTransposeWithDynamicPadsHelper : public ConvolutionHelperBase
+{
+public:
+    template<typename Info_t, typename Shape_t>
+    ConvTransposeWithDynamicPadsHelper(const Info_t& info, const Shape_t& shape) : ConvolutionHelperBase(info, shape, true, true) {}
 };
 
 class GemmHelper
@@ -501,20 +534,138 @@ class SplitHelper
     std::vector<int> m_split;
 };
 
-class SliceHelper
+class SliceHelperBase
 {
 public:
+    template<typename Info_t, typename Index_t>
+    void ReadIndexTensors(
+        const Info_t& operatorInfo,
+        std::vector<int32_t>& starts,
+        std::vector<int32_t>& ends,
+        std::vector<int32_t>& axes,
+        std::vector<int32_t>& steps
+    )
+    {
+        // Get starts, ends, optional axes and optional steps from constant inputs.
+        MLOperatorTensor startsTensor = operatorInfo.GetConstantInputTensor(1);
+        const std::vector<uint32_t>& startsTensorDimensions = startsTensor.GetShape();
+        size_t dimCount = startsTensorDimensions[0];
+        const Index_t* startsData = startsTensor.GetData<Index_t>();
+        for (size_t i = 0; i < dimCount; ++i)
+        {
+            starts.push_back(gsl::narrow_cast<int32_t>(startsData[i]));
+        }
+
+        MLOperatorTensor endsTensor = operatorInfo.GetConstantInputTensor(2);
+        const std::vector<uint32_t>& endsTensorDimensions = endsTensor.GetShape();
+        dimCount = endsTensorDimensions[0];
+        const Index_t* endsData = endsTensor.GetData<Index_t>();
+        for (size_t i = 0; i < dimCount; ++i)
+        {
+            ends.push_back(gsl::narrow_cast<int32_t>(endsData[i]));
+        }
+        uint32_t inputCount = operatorInfo.GetInputCount();
+        if (operatorInfo.GetInputCount() > 3)
+        {
+            MLOperatorTensor axesTensor = operatorInfo.GetConstantInputTensor(3);
+            const std::vector<uint32_t>& axesTensorDimensions = axesTensor.GetShape();
+            dimCount = axesTensorDimensions[0];
+            const Index_t* axesData = axesTensor.GetData<Index_t>();
+            for (size_t i = 0; i < dimCount; ++i)
+            {
+                axes.push_back(gsl::narrow_cast<int32_t>(axesData[i]));
+            }
+        }
+
+        if (operatorInfo.GetInputCount() > 4)
+        {
+            MLOperatorTensor stepsTensor = operatorInfo.GetConstantInputTensor(4);
+            const std::vector<uint32_t>& stepsTensorDimensions = stepsTensor.GetShape();
+            dimCount = stepsTensorDimensions[0];
+            const Index_t* stepsData = stepsTensor.GetData<Index_t>();
+            for (size_t i = 0; i < dimCount; ++i)
+            {
+                steps.push_back(gsl::narrow_cast<int32_t>(stepsData[i]));
+            }
+        }
+    }
+
+    template<typename Info_t>
     void Initialize(
-        const MLOperatorAttributes& operatorAttributes,
-        gsl::span<const DimensionType> inputDimensions
-        );
+        const Info_t& operatorInfo,
+        gsl::span<const DimensionType> inputDimensions,
+        uint32_t opsetVersion
+        )
+    {
+        std::vector<int32_t> starts;
+        std::vector<int32_t> ends;
+        std::vector<int32_t> axes;
+        std::vector<int32_t> steps;
+        if (opsetVersion == 7)
+        {
+            // Get starts, ends and axes from attributes
+            starts = operatorInfo.GetOptionalAttributeVectorInt32(AttrName::Starts);
+            ends = operatorInfo.GetOptionalAttributeVectorInt32(AttrName::Ends);
+            axes = operatorInfo.GetOptionalAttributeVectorInt32(AttrName::Axes);
+        }
+        else if (opsetVersion == 10)
+        {
+            if (operatorInfo.GetConstantInputTensor(1).GetTensorDataType() == MLOperatorTensorDataType::Int32)
+            {
+                ReadIndexTensors<Info_t, int32_t>(operatorInfo, starts, ends, axes, steps);
+            }
+            else
+            {
+                THROW_HR_IF(E_INVALIDARG, operatorInfo.GetConstantInputTensor(1).GetTensorDataType() != MLOperatorTensorDataType::Int64);
+                ReadIndexTensors<Info_t, int64_t>(operatorInfo, starts, ends, axes, steps);
+            }
+        }
+         
+        ML_CHECK_VALID_ARGUMENT(starts.size() == ends.size(), "'starts' must equal 'ends' in size.");
+        ML_CHECK_VALID_ARGUMENT(axes.empty() || starts.size() == axes.size(), "'axes' must equal 'starts' in size, or 'axes' must be empty.");
+
+        m_outputDimensions.assign(inputDimensions.begin(), inputDimensions.end());
+        m_offsets.resize(m_outputDimensions.size());
+        m_sizes.resize(m_outputDimensions.size());
+        m_strides.resize(m_outputDimensions.size(), 1); // Only a stride of 1 element is supported by ONNX 1.2.
+
+        // Set initial defaults lest 'starts' and 'ends' arrays are shorter than the dimension count.
+        std::copy(inputDimensions.begin(), inputDimensions.begin() + m_outputDimensions.size(), m_sizes.begin());
+
+        // Clamp selected dimensions to given 'starts' and 'ends'.
+        for (int i = 0, ci = gsl::narrow_cast<int>(starts.size()); i < ci; ++i)
+        {
+            int dimIndex = i;
+            if (!axes.empty())
+            {
+                dimIndex = axes[i];
+            }
+            ML_CHECK_VALID_ARGUMENT(dimIndex < inputDimensions.size(), "'axes' must be valid with within actual input dimensions.");
+
+            // Positive values are offsets from 0.
+            // Negative values are offsets from the dimension's size.
+            int dim = gsl::narrow_cast<int>(inputDimensions[dimIndex]);
+            int start = (starts[i] < 0) ? (starts[i] + dim) : starts[i];
+            int end = (ends[i] < 0) ? (ends[i] + dim) : ends[i];
+
+            // Clamp the dimensions to the slice extents.
+            // Clamp negative numbers to 0, per case test_slice_start_out_of_bounds.
+            start = std::max(start, 0);
+            end = std::min(end, dim);
+            int size = std::max(end - start, 0);
+
+            m_outputDimensions[dimIndex] = size;
+            m_offsets[dimIndex] = start;
+            m_sizes[dimIndex] = gsl::narrow_cast<uint32_t>(size);
+        }
+    }
 
     // Info_t is used to obtain attributes which will be used for calculating the output shape later. 
     // Shape_t is used to obtain input shape which will be used for adjusting attribute value. 
     template<typename Info_t, typename Shape_t>
-    SliceHelper(const Info_t& info, const Shape_t& shape)
+    SliceHelperBase(const Info_t& info, const Shape_t& shape, uint32_t opsetVersion)
     {
-        Initialize(info, shape.GetInputTensorShape(0));
+        Initialize(info, shape.GetInputTensorShape(0), opsetVersion);
     }
 
     std::vector<EdgeShapes> GetOutputShapes(const MLShapeInferenceContext& shapeInfo) const;
@@ -526,6 +677,21 @@ class SliceHelper
     std::vector<uint32_t> m_strides;
 };
 
+class SliceHelper : public SliceHelperBase
+{
+public:
+    template<typename Info_t, typename Shape_t>
+    SliceHelper(const Info_t& info, const Shape_t& shape) : SliceHelperBase(info, shape, 7) {}
+};
+
+class Slice10Helper : public SliceHelperBase
+{
+public:
+    template<typename Info_t, typename Shape_t>
+    Slice10Helper(const Info_t& info, const Shape_t& shape) : SliceHelperBase(info, shape, 10) {}
+};
+
+
 class PaddingHelper
 {
 public:
@@ -1098,6 +1264,7 @@ class OneHotHelper
 
 using ShapeInferenceHelper_Conv = ConvHelper;
 using ShapeInferenceHelper_ConvTranspose = ConvTransposeHelper;
+using ShapeInferenceHelper_ConvTransposeWithDynamicPads = ConvTransposeWithDynamicPadsHelper;
 using ShapeInferenceHelper_AveragePool = PoolingHelper;
 using ShapeInferenceHelper_GlobalAveragePool = GlobalPoolingHelper;
 using ShapeInferenceHelper_MaxPool = PoolingHelper;
@@ -1120,7 +1287,8 @@ using ShapeInferenceHelper_Flatten = FlattenHelper;
 using ShapeInferenceHelper_Split = SplitHelper;
 using ShapeInferenceHelper_Transpose = TransposeHelper;
 using ShapeInferenceHelper_Concat = ConcatHelper;
-using ShapeInferenceHelper_Slice = SliceHelper;
+using ShapeInferenceHelper_Slice7 = SliceHelper;
+using ShapeInferenceHelper_Slice10 = Slice10Helper;
 using ShapeInferenceHelper_Pad = PaddingHelper;
 using ShapeInferenceHelper_SpaceToDepth = SpaceToDepthHelper;
 using ShapeInferenceHelper_DepthToSpace = DepthToSpaceHelper;
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorRegistration.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorRegistration.h
index b9e429a6bf1bd..2b28bf9f6209b 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorRegistration.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorRegistration.h
@@ -178,6 +178,7 @@ namespace OperatorHelper
         static const int sc_sinceVer_Dropout = 10;
         static const int sc_sinceVer_ThresholdedRelu = 10;
         static const int sc_sinceVer_Upsample = 10;
+        static const int sc_sinceVer_Slice = 10;
     } // namespace OnnxOperatorSet10
 
     namespace MsftOperatorSet1
@@ -193,6 +194,7 @@ namespace OperatorHelper
         static const int sc_sinceVer_FusedSum = 1;
         static const int sc_sinceVer_QuantizeLinear = 1;
         static const int sc_sinceVer_DequantizeLinear = 1;
+        static const int sc_sinceVer_ConvTransposeWithDynamicPads = 1;
     } // namespace MsftOperatorSet1
 
 } // namespace OperatorHelper
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index ed444ccc67ae2..d8bd447e099aa 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -563,7 +563,6 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
     broken_tests.insert({"dynamicquantizelinear_expanded", "Temporarily disabled pending investigation"});
     broken_tests.insert({"dynamicquantizelinear_max_adjusted_expanded", "Temporarily disabled pending investigation"});
     broken_tests.insert({"dynamicquantizelinear_min_adjusted_expanded", "Temporarily disabled pending investigation"});
-    broken_tests.insert({"maxpool_with_argmax_2d_precomputed_pads", "Temporarily disabled pending investigation"});
     broken_tests.insert({"mxnet_arcface", "Temporarily disabled pending investigation"});
     broken_tests.insert({"yolov3", "Temporarily disabled pending investigation"});
     broken_tests.insert({"tf_inception_v2", "Temporarily disabled pending investigation"});

From 71940c09157399beabec57b545d81047b369bbd0 Mon Sep 17 00:00:00 2001
From: KeDengMS <kedeng@microsoft.com>
Date: Sun, 22 Dec 2019 22:42:03 -0800
Subject: [PATCH 05/75] Update Nuphar tutorial notebook (#2721)

1. Reflect int8 GEMV improvements for multi-threading from #2696
2. Add notes on multi-threading control using OpenMP
3. Add samples of running multi-isa AOT, and show int8 GEMM differences between AVX and AVX2
4. Add rnn_benchmark example to resolve #1993
---
 .../onnxruntime-nuphar-tutorial.ipynb         | 445 +++++++++++-------
 1 file changed, 280 insertions(+), 165 deletions(-)

diff --git a/docs/python/notebooks/onnxruntime-nuphar-tutorial.ipynb b/docs/python/notebooks/onnxruntime-nuphar-tutorial.ipynb
index 2515dfed0bb45..2872f35781c9a 100644
--- a/docs/python/notebooks/onnxruntime-nuphar-tutorial.ipynb
+++ b/docs/python/notebooks/onnxruntime-nuphar-tutorial.ipynb
@@ -46,6 +46,7 @@
    "outputs": [],
    "source": [
     "import cpufeature\n",
+    "import hashlib\n",
     "import numpy as np\n",
     "import onnx\n",
     "from onnx import helper, numpy_helper\n",
@@ -67,7 +68,20 @@
     "\n",
     "def print_speedup(name, delta_baseline, delta):\n",
     "    print(\"{} speed-up {:.2f}%\".format(name, 100*(delta_baseline/delta - 1)))\n",
-    "    print(\"    Baseline: {:.3f} s, Current: {:.3f} s\".format(delta_baseline, delta))"
+    "    print(\"    Baseline: {:.3f} s, Current: {:.3f} s\".format(delta_baseline, delta))\n",
+    "\n",
+    "def create_cache_dir(cache_dir):\n",
+    "    # remove any stale cache files\n",
+    "    if os.path.exists(cache_dir):\n",
+    "        shutil.rmtree(cache_dir)\n",
+    "    os.makedirs(cache_dir, exist_ok=True)\n",
+    "\n",
+    "def md5(file_name):\n",
+    "    hash_md5 = hashlib.md5()\n",
+    "    with open(file_name, \"rb\") as f:\n",
+    "        for chunk in iter(lambda: f.read(4096), b\"\"):\n",
+    "            hash_md5.update(chunk)\n",
+    "    return hash_md5.hexdigest()"
    ]
   },
   {
@@ -86,7 +100,7 @@
     "import onnxruntime\n",
     "from onnxruntime.nuphar.model_editor import convert_to_scan_model\n",
     "from onnxruntime.nuphar.model_quantizer import convert_matmul_model\n",
-    "from onnxruntime.nuphar.rnn_benchmark import generate_model\n",
+    "from onnxruntime.nuphar.rnn_benchmark import generate_model, perf_test\n",
     "from onnxruntime.nuphar.symbolic_shape_infer import SymbolicShapeInference"
    ]
   },
@@ -216,8 +230,8 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Fusion speed-up 436.78%\n",
-      "    Baseline: 0.725 s, Current: 0.135 s\n"
+      "Fusion speed-up 445.44%\n",
+      "    Baseline: 0.732 s, Current: 0.134 s\n"
      ]
     }
    ],
@@ -226,18 +240,18 @@
     "batch = 16\n",
     "input_data = np.random.rand(seq, batch, dim).astype(np.float32)\n",
     "sess = onnxruntime.InferenceSession(simple_model_name)\n",
-    "feed = {X:input_data}\n",
-    "output = sess.run([], feed)\n",
+    "simple_feed = {X:input_data}\n",
+    "simple_output = sess.run([], simple_feed)\n",
     "np_output = ((((input_data + input_data) * input_data) + input_data) * input_data) + input_data\n",
-    "assert np.allclose(output[0], np_output)\n",
+    "assert np.allclose(simple_output[0], np_output)\n",
     "\n",
-    "repeats = 100\n",
+    "simple_repeats = 100\n",
     "start_ort = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess.run([], feed)\n",
+    "for i in range(simple_repeats):\n",
+    "    sess.run([], simple_feed)\n",
     "end_ort = timer()\n",
     "start_np = timer()\n",
-    "for i in range(repeats):\n",
+    "for i in range(simple_repeats):\n",
     "    np_output = ((((input_data + input_data) * input_data) + input_data) * input_data) + input_data\n",
     "end_np = timer()\n",
     "print_speedup('Fusion', end_np - start_np, end_ort - start_ort)"
@@ -271,7 +285,9 @@
     "collapsed": true
    },
    "source": [
-    "**IMPORTANT**: Nuphar generates code before knowing shapes of input data, unlike other execution providers that do runtime shape inference. Thus, shape inference information is critical for compiler optimizations in Nuphar. To do that, we run symbolic shape inference on the model. Symbolic shape inference is based on the ONNX shape inference, and enhanced by sympy to better handle Shape/ConstantOfShape/etc. ops using symbolic computation."
+    "**IMPORTANT**: Nuphar generates code before knowing shapes of input data, unlike other execution providers that do runtime shape inference. Thus, shape inference information is critical for compiler optimizations in Nuphar. To do that, we run symbolic shape inference on the model. Symbolic shape inference is based on the ONNX shape inference, and enhanced by sympy to better handle Shape/ConstantOfShape/etc. ops using symbolic computation.\n",
+    "\n",
+    "**IMPORTANT**: When running multi-threaded inference, Nuphar currently uses TVM's parallel schedule with has its own thread pool that's compatible with OpenMP and MKLML. The TVM thread pool has not been integrated with ONNX runtime thread pool, so intra_op_num_threads won't control it. Please make sure the build is with OpenMP or MKLML, and use OMP_NUM_THREADS to control thread pool."
    ]
   },
   {
@@ -298,12 +314,11 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "sess_baseline = onnxruntime.InferenceSession(lstm_model)\n",
-    "sess_baseline.set_providers(['CPUExecutionProvider']) # default provider in this container is Nuphar, this overrides to CPU EP\n",
+    "sess_baseline = onnxruntime.InferenceSession(lstm_model, providers=['CPUExecutionProvider'])\n",
     "seq = 128\n",
     "input_data = np.random.rand(seq, 1, input_dim).astype(np.float32)\n",
-    "feed = {sess_baseline.get_inputs()[0].name:input_data}\n",
-    "output = sess_baseline.run([], feed)"
+    "lstm_feed = {sess_baseline.get_inputs()[0].name:input_data}\n",
+    "lstm_output = sess_baseline.run([], lstm_feed)"
    ]
   },
   {
@@ -319,8 +334,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "scan_model = 'Scan_LSTMx4.onnx'\n",
-    "convert_to_scan_model(lstm_model, scan_model)"
+    "lstm_scan_model = 'Scan_LSTMx4.onnx'\n",
+    "convert_to_scan_model(lstm_model, lstm_scan_model)"
    ]
   },
   {
@@ -339,28 +354,28 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Nuphar Scan speed-up 3.37%\n",
-      "    Baseline: 3.099 s, Current: 2.998 s\n"
+      "Nuphar Scan speed-up 9.22%\n",
+      "    Baseline: 3.070 s, Current: 2.811 s\n"
      ]
     }
    ],
    "source": [
-    "sess_nuphar = onnxruntime.InferenceSession(scan_model)\n",
-    "output_nuphar = sess_nuphar.run([], feed)\n",
-    "assert np.allclose(output[0], output_nuphar[0])\n",
+    "sess_nuphar = onnxruntime.InferenceSession(lstm_scan_model)\n",
+    "output_nuphar = sess_nuphar.run([], lstm_feed)\n",
+    "assert np.allclose(lstm_output[0], output_nuphar[0])\n",
     "\n",
-    "repeats = 10\n",
-    "start_baseline = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_baseline.run([], feed)\n",
-    "end_baseline = timer()\n",
+    "lstm_repeats = 10\n",
+    "start_lstm_baseline = timer()\n",
+    "for i in range(lstm_repeats):\n",
+    "    sess_baseline.run([], lstm_feed)\n",
+    "end_lstm_baseline = timer()\n",
     "\n",
     "start_nuphar = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_nuphar.run([], feed)\n",
+    "for i in range(lstm_repeats):\n",
+    "    sess_nuphar.run([], lstm_feed)\n",
     "end_nuphar = timer()\n",
     "\n",
-    "print_speedup('Nuphar Scan', end_baseline - start_baseline, end_nuphar - start_nuphar)"
+    "print_speedup('Nuphar Scan', end_lstm_baseline - start_lstm_baseline, end_nuphar - start_nuphar)"
    ]
   },
   {
@@ -406,8 +421,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "quantized_model = 'Scan_LSTMx4_int8.onnx'\n",
-    "convert_matmul_model(scan_model, quantized_model)"
+    "lstm_quantized_model = 'Scan_LSTMx4_int8.onnx'\n",
+    "convert_matmul_model(lstm_scan_model, lstm_quantized_model)"
    ]
   },
   {
@@ -423,9 +438,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "sess_quantized = onnxruntime.InferenceSession(quantized_model)\n",
-    "output_quantized = sess_quantized.run([], feed)\n",
-    "assert np.allclose(output[0], output_quantized[0], rtol=1e-3, atol=1e-3)"
+    "sess_quantized = onnxruntime.InferenceSession(lstm_quantized_model)\n",
+    "output_quantized = sess_quantized.run([], lstm_feed)\n",
+    "assert np.allclose(lstm_output[0], output_quantized[0], rtol=1e-3, atol=1e-3)"
    ]
   },
   {
@@ -444,27 +459,64 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Quantization speed-up 299.66%\n",
-      "    Baseline: 2.998 s, Current: 0.750 s\n"
+      "Quantization speed-up 722.57%\n",
+      "    Baseline: 2.811 s, Current: 0.342 s\n"
      ]
     }
    ],
    "source": [
     "start_quantized = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_quantized.run([], feed)\n",
+    "for i in range(lstm_repeats):\n",
+    "    sess_quantized.run([], lstm_feed)\n",
     "end_quantized = timer()\n",
     "\n",
     "print_speedup('Quantization', end_nuphar - start_nuphar, end_quantized - start_quantized)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "To check RNN quantization performance, please use rnn_benchmark.perf_test."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "perf_rnn (with default threads) lstm_i80_h512_bi_l6_no_batch.onnx: run for 348 iterations, top 5 avg 28.368 ms\n",
+      "perf_scan (with 4 threads) lstm_i80_h512_bi_l6_no_batch_scan.onnx: run for 373 iterations, top 5 avg 25.819 ms\n",
+      "perf_int8 (with 4 threads) lstm_i80_h512_bi_l6_no_batch_int8.onnx: run for 778 iterations, top 5 avg 12.534 ms\n",
+      "Nuphar Quantization speed up speed-up 126.33%\n",
+      "    Baseline: 0.028 s, Current: 0.013 s\n"
+     ]
+    }
+   ],
+   "source": [
+    "rnn_type = 'lstm'                         # could be 'lstm', 'gru' or 'rnn'\n",
+    "num_threads = cpufeature.CPUFeature['num_physical_cores'] # no hyper thread\n",
+    "input_dim = 80                            # size of input dimension\n",
+    "hidden_dim = 512                          # size of hidden dimension in cell\n",
+    "bidirectional = True                      # specify RNN being bidirectional\n",
+    "layers = 6                                # number of stacked RNN layers\n",
+    "seq_len = 40                              # length of sequence\n",
+    "batch_size = 1                            # size of batch\n",
+    "original_ms, scan_ms, int8_ms = perf_test(rnn_type, num_threads, input_dim, hidden_dim, bidirectional, layers, seq_len, batch_size)\n",
+    "print_speedup('Nuphar Quantization speed up', original_ms / 1000, int8_ms / 1000)"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
     "## 5. Working on real models\n",
     "\n",
-    "### BERT Squad\n",
+    "### 5.1 BERT Squad\n",
     "\n",
     "BERT (Bidirectional Encoder Representations from Transformers) applies Transformers to language modelling. With Nuphar, we may fuse and compile the model to accelerate inference on CPU.\n",
     "\n",
@@ -473,17 +525,17 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 16,
+   "execution_count": 17,
    "metadata": {},
    "outputs": [],
    "source": [
     "# download BERT squad model\n",
     "cwd = os.getcwd()\n",
-    "model_url = 'https://onnxzoo.blob.core.windows.net/models/opset_10/bert_squad/download_sample_10.tar.gz'\n",
-    "model_local = os.path.join(cwd, 'download_sample_10.tar.gz')\n",
-    "if not os.path.exists(model_local):\n",
-    "  urllib.request.urlretrieve(model_url, model_local)\n",
-    "with tarfile.open(model_local, 'r') as f:\n",
+    "bert_model_url = 'https://onnxzoo.blob.core.windows.net/models/opset_10/bert_squad/download_sample_10.tar.gz'\n",
+    "bert_model_local = os.path.join(cwd, 'download_sample_10.tar.gz')\n",
+    "if not os.path.exists(bert_model_local):\n",
+    "  urllib.request.urlretrieve(bert_model_url, bert_model_local)\n",
+    "with tarfile.open(bert_model_local, 'r') as f:\n",
     "  f.extractall(cwd)"
    ]
   },
@@ -497,16 +549,16 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [],
    "source": [
-    "model_dir = os.path.join(cwd, 'download_sample_10')\n",
-    "model = os.path.join(model_dir, 'bertsquad10.onnx')\n",
-    "model_with_shape_inference = os.path.join(model_dir, 'bertsquad10_shaped.onnx')\n",
+    "bert_model_dir = os.path.join(cwd, 'download_sample_10')\n",
+    "bert_model = os.path.join(bert_model_dir, 'bertsquad10.onnx')\n",
+    "bert_model_with_shape_inference = os.path.join(bert_model_dir, 'bertsquad10_shaped.onnx')\n",
     "\n",
     "# run symbolic shape inference\n",
-    "SymbolicShapeInference.infer_shapes(model, model_with_shape_inference, auto_merge=True, int_max=100000)"
+    "SymbolicShapeInference.infer_shapes(bert_model, bert_model_with_shape_inference, auto_merge=True, int_max=100000)"
    ]
   },
   {
@@ -518,26 +570,25 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 18,
+   "execution_count": 19,
    "metadata": {},
    "outputs": [],
    "source": [
     "sess_options = onnxruntime.SessionOptions()\n",
     "sess_options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL\n",
-    "sess_baseline = onnxruntime.InferenceSession(model, sess_options)\n",
-    "sess_baseline.set_providers(['CPUExecutionProvider'])\n",
+    "sess_baseline = onnxruntime.InferenceSession(bert_model, sess_options=sess_options, providers=['CPUExecutionProvider'])\n",
     "\n",
     "# load test data\n",
-    "test_data_dir = os.path.join(model_dir, 'test_data_set_1')\n",
+    "test_data_dir = os.path.join(bert_model_dir, 'test_data_set_1')\n",
     "tps = [onnx.load_tensor(os.path.join(test_data_dir, 'input_{}.pb'.format(i))) for i in range(len(sess_baseline.get_inputs()))]\n",
-    "feed = {tp.name:numpy_helper.to_array(tp) for tp in tps}\n",
-    "output_baseline = sess_baseline.run([], feed)\n",
+    "bert_feed = {tp.name:numpy_helper.to_array(tp) for tp in tps}\n",
+    "bert_output_baseline = sess_baseline.run([], bert_feed)\n",
     "\n",
-    "repeats = 20\n",
-    "start_baseline = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_baseline.run([], feed)\n",
-    "end_baseline = timer()"
+    "bert_repeats = 20\n",
+    "start_bert_baseline = timer()\n",
+    "for i in range(bert_repeats):\n",
+    "    sess_baseline.run([], bert_feed)\n",
+    "end_bert_baseline = timer()"
    ]
   },
   {
@@ -550,13 +601,13 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
+   "execution_count": 20,
    "metadata": {},
    "outputs": [],
    "source": [
-    "sess = onnxruntime.InferenceSession(model_with_shape_inference)\n",
-    "output = sess.run([], feed)\n",
-    "assert all([np.allclose(o, ob, atol=1e-4) for o, ob in zip(output, output_baseline)])"
+    "sess = onnxruntime.InferenceSession(bert_model_with_shape_inference)\n",
+    "output = sess.run([], bert_feed)\n",
+    "assert all([np.allclose(o, ob, atol=1e-4) for o, ob in zip(output, bert_output_baseline)])"
    ]
   },
   {
@@ -568,32 +619,32 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 21,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Nuphar BERT squad speed-up 60.13%\n",
-      "    Baseline: 4.928 s, Current: 3.077 s\n"
+      "Nuphar BERT squad speed-up 66.70%\n",
+      "    Baseline: 5.093 s, Current: 3.055 s\n"
      ]
     }
    ],
    "source": [
     "start_nuphar = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess.run([], feed)\n",
+    "for i in range(bert_repeats):\n",
+    "    sess.run([], bert_feed)\n",
     "end_nuphar = timer()\n",
     "\n",
-    "print_speedup('Nuphar BERT squad', end_baseline - start_baseline, end_nuphar - start_nuphar)"
+    "print_speedup('Nuphar BERT squad', end_bert_baseline - start_bert_baseline, end_nuphar - start_nuphar)"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "### GPT-2 with fixed batch size\n",
+    "### 5.2 GPT-2 with fixed batch size\n",
     "GPT-2 is a language model using Generative Pre-Trained Transformer for text generation. With Nuphar, we may fuse and compile the model to accelerate inference on CPU.\n",
     "\n",
     "#### Download model and test data"
@@ -601,17 +652,17 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 21,
+   "execution_count": 22,
    "metadata": {},
    "outputs": [],
    "source": [
     "# download GPT-2 model\n",
     "cwd = os.getcwd()\n",
-    "model_url = 'https://onnxzoo.blob.core.windows.net/models/opset_10/GPT2/GPT-2.tar.gz'\n",
-    "model_local = os.path.join(cwd, 'GPT-2.tar.gz')\n",
-    "if not os.path.exists(model_local):\n",
-    "  urllib.request.urlretrieve(model_url, model_local)\n",
-    "with tarfile.open(model_local, 'r') as f:\n",
+    "gpt2_model_url = 'https://onnxzoo.blob.core.windows.net/models/opset_10/GPT2/GPT-2.tar.gz'\n",
+    "gpt2_model_local = os.path.join(cwd, 'GPT-2.tar.gz')\n",
+    "if not os.path.exists(gpt2_model_local):\n",
+    "  urllib.request.urlretrieve(gpt2_model_url, gpt2_model_local)\n",
+    "with tarfile.open(gpt2_model_local, 'r') as f:\n",
     "  f.extractall(cwd)"
    ]
   },
@@ -625,22 +676,22 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": 23,
    "metadata": {},
    "outputs": [],
    "source": [
-    "model_dir = os.path.join(cwd, 'GPT2')\n",
-    "model = os.path.join(model_dir, 'model.onnx')\n",
+    "gpt2_model_dir = os.path.join(cwd, 'GPT2')\n",
+    "gpt2_model = os.path.join(gpt2_model_dir, 'model.onnx')\n",
     "\n",
     "# edit batch dimension from symbolic to int value for better codegen\n",
-    "mp = onnx.load(model)\n",
+    "mp = onnx.load(gpt2_model)\n",
     "mp.graph.input[0].type.tensor_type.shape.dim[0].dim_value = 1\n",
-    "onnx.save(mp, model)\n",
+    "onnx.save(mp, gpt2_model)\n",
     "\n",
-    "model_with_shape_inference = os.path.join(model_dir, 'model_shaped.onnx')\n",
+    "gpt2_model_with_shape_inference = os.path.join(gpt2_model_dir, 'model_shaped.onnx')\n",
     "\n",
     "# run symbolic shape inference\n",
-    "SymbolicShapeInference.infer_shapes(model, model_with_shape_inference, auto_merge=True)"
+    "SymbolicShapeInference.infer_shapes(gpt2_model, gpt2_model_with_shape_inference, auto_merge=True)"
    ]
   },
   {
@@ -652,54 +703,53 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 23,
+   "execution_count": 24,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Nuphar GPT-2 speed-up 13.48%\n",
-      "    Baseline: 2.535 s, Current: 2.234 s\n"
+      "Nuphar GPT-2 speed-up 16.15%\n",
+      "    Baseline: 2.671 s, Current: 2.299 s\n"
      ]
     }
    ],
    "source": [
     "sess_options = onnxruntime.SessionOptions()\n",
     "sess_options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL\n",
-    "sess_baseline = onnxruntime.InferenceSession(model, sess_options)\n",
-    "sess_baseline.set_providers(['CPUExecutionProvider'])\n",
+    "sess_baseline = onnxruntime.InferenceSession(gpt2_model, sess_options=sess_options, providers=['CPUExecutionProvider'])\n",
     "\n",
     "# load test data, note the tensor proto name in data does not match model, so override it in feed\n",
     "input_name = [i.name for i in sess_baseline.get_inputs()][0] # This model only has one input\n",
-    "test_data_dir = os.path.join(model_dir, 'test_data_set_0')\n",
+    "test_data_dir = os.path.join(gpt2_model_dir, 'test_data_set_0')\n",
     "tp = onnx.load_tensor(os.path.join(test_data_dir, 'input_0.pb'))\n",
-    "feed = {input_name:numpy_helper.to_array(tp).reshape(1,-1)} # the test data missed batch dimension\n",
-    "output_baseline = sess_baseline.run([], feed)\n",
+    "gpt2_feed = {input_name:numpy_helper.to_array(tp).reshape(1,-1)} # the test data missed batch dimension\n",
+    "gpt2_output_baseline = sess_baseline.run([], gpt2_feed)\n",
     "\n",
-    "repeats = 100\n",
-    "start_baseline = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_baseline.run([], feed)\n",
-    "end_baseline = timer()\n",
+    "gpt2_repeats = 100\n",
+    "start_gpt2_baseline = timer()\n",
+    "for i in range(gpt2_repeats):\n",
+    "    sess_baseline.run([], gpt2_feed)\n",
+    "end_gpt2_baseline = timer()\n",
     "\n",
-    "sess = onnxruntime.InferenceSession(model_with_shape_inference)\n",
-    "output = sess.run([], feed)\n",
-    "assert all([np.allclose(o, ob, atol=1e-4) for o, ob in zip(output, output_baseline)])\n",
+    "sess = onnxruntime.InferenceSession(gpt2_model_with_shape_inference)\n",
+    "output = sess.run([], gpt2_feed)\n",
+    "assert all([np.allclose(o, ob, atol=1e-4) for o, ob in zip(output, gpt2_output_baseline)])\n",
     "\n",
     "start_nuphar = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess.run([], feed)\n",
+    "for i in range(gpt2_repeats):\n",
+    "    output = sess.run([], gpt2_feed)\n",
     "end_nuphar = timer()\n",
     "\n",
-    "print_speedup('Nuphar GPT-2', end_baseline - start_baseline, end_nuphar - start_nuphar)"
+    "print_speedup('Nuphar GPT-2', end_gpt2_baseline - start_gpt2_baseline, end_nuphar - start_nuphar)"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "### BiDAF with quantization\n",
+    "### 5.3 BiDAF with quantization\n",
     "\n",
     "BiDAF is a machine comprehension model that uses LSTMs. The inputs to this model are paragraphs of contexts and queries, and the outputs are start/end indices of words in the contexts that answers the queries.\n",
     "\n",
@@ -708,7 +758,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 24,
+   "execution_count": 25,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -731,18 +781,18 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 25,
+   "execution_count": 26,
    "metadata": {},
    "outputs": [],
    "source": [
-    "bidaf = os.path.join(cwd, 'bidaf', 'bidaf.onnx')\n",
-    "sess_baseline = onnxruntime.InferenceSession(bidaf)\n",
-    "sess_baseline.set_providers(['CPUExecutionProvider'])\n",
+    "bidaf_dir = os.path.join(cwd, 'bidaf')\n",
+    "bidaf = os.path.join(bidaf_dir, 'bidaf.onnx')\n",
+    "sess_baseline = onnxruntime.InferenceSession(bidaf, providers=['CPUExecutionProvider'])\n",
     "# load test data\n",
     "test_data_dir = os.path.join(cwd, 'bidaf', 'test_data_set_3')\n",
     "tps = [onnx.load_tensor(os.path.join(test_data_dir, 'input_{}.pb'.format(i))) for i in range(len(sess_baseline.get_inputs()))]\n",
-    "feed = {tp.name:numpy_helper.to_array(tp) for tp in tps}\n",
-    "output_baseline = sess_baseline.run([], feed)"
+    "bidaf_feed = {tp.name:numpy_helper.to_array(tp) for tp in tps}\n",
+    "bidaf_output_baseline = sess_baseline.run([], bidaf_feed)"
    ]
   },
   {
@@ -754,7 +804,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 26,
+   "execution_count": 27,
    "metadata": {},
    "outputs": [
     {
@@ -763,13 +813,13 @@
        "\"with 4:51 left in regulation , carolina got the ball on their own 24 - yard line with a chance to mount a game - winning drive , and soon faced 3rd - and - 9 . on the next play , miller stripped the ball away from newton , and after several players dove for it , it took a long bounce backwards and was recovered by ward , who returned it five yards to the panthers 4 - yard line . although several players dove into the pile to attempt to recover it , newton did not and his lack of aggression later earned him heavy criticism . meanwhile , denver  ' s offense was kept out of the end zone for three plays , but a holding penalty on cornerback josh norman gave the broncos a new set of downs . then anderson scored on a 2 - yard touchdown run and manning completed a pass to bennie fowler for a 2 - point conversion , giving denver a 24 – 10 lead with 3:08 left and essentially putting the game away . carolina had two more drives , but failed to get a first down on each one .\""
       ]
      },
-     "execution_count": 26,
+     "execution_count": 27,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "' '.join(list(feed['context_word'].reshape(-1)))"
+    "' '.join(list(bidaf_feed['context_word'].reshape(-1)))"
    ]
   },
   {
@@ -781,7 +831,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 27,
+   "execution_count": 28,
    "metadata": {},
    "outputs": [
     {
@@ -790,13 +840,13 @@
        "'who recovered the strip ball ?'"
       ]
      },
-     "execution_count": 27,
+     "execution_count": 28,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "' '.join(list(feed['query_word'].reshape(-1)))"
+    "' '.join(list(bidaf_feed['query_word'].reshape(-1)))"
    ]
   },
   {
@@ -808,7 +858,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": 29,
    "metadata": {},
    "outputs": [
     {
@@ -817,13 +867,13 @@
        "'ward'"
       ]
      },
-     "execution_count": 28,
+     "execution_count": 29,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "' '.join(list(feed['context_word'][output_baseline[0][0]:output_baseline[1][0]+1].reshape(-1)))"
+    "' '.join(list(bidaf_feed['context_word'][bidaf_output_baseline[0][0]:bidaf_output_baseline[1][0]+1].reshape(-1)))"
    ]
   },
   {
@@ -835,7 +885,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": 30,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -851,8 +901,8 @@
     "\n",
     "# inference and verify accuracy\n",
     "sess = onnxruntime.InferenceSession(bidaf_converted)\n",
-    "output = sess.run([], feed)\n",
-    "assert all([np.allclose(o, ob) for o, ob in zip(output, output_baseline)])"
+    "output = sess.run([], bidaf_feed)\n",
+    "assert all([np.allclose(o, ob) for o, ob in zip(output, bidaf_output_baseline)])"
    ]
   },
   {
@@ -864,30 +914,31 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": 31,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Nuphar quantized BiDAF speed-up 47.77%\n",
-      "    Baseline: 1.564 s, Current: 1.058 s\n"
+      "Nuphar quantized BiDAF speed-up 43.79%\n",
+      "    Baseline: 1.517 s, Current: 1.055 s\n"
      ]
     }
    ],
    "source": [
-    "start_baseline = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_baseline.run([], feed)\n",
-    "end_baseline = timer()\n",
+    "bidaf_repeats = 100\n",
+    "start_bidaf_baseline = timer()\n",
+    "for i in range(bidaf_repeats):\n",
+    "    sess_baseline.run([], bidaf_feed)\n",
+    "end_bidaf_baseline = timer()\n",
     "\n",
     "start_nuphar = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess.run([], feed)\n",
+    "for i in range(bidaf_repeats):\n",
+    "    sess.run([], bidaf_feed)\n",
     "end_nuphar = timer()\n",
     "\n",
-    "print_speedup('Nuphar quantized BiDAF', end_baseline - start_baseline, end_nuphar - start_nuphar)"
+    "print_speedup('Nuphar quantized BiDAF', end_bidaf_baseline - start_bidaf_baseline, end_nuphar - start_nuphar)"
    ]
   },
   {
@@ -907,16 +958,16 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 31,
+   "execution_count": 32,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "'JIT took 4.721 seconds'"
+       "'JIT took 5.749 seconds'"
       ]
      },
-     "execution_count": 31,
+     "execution_count": 32,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -930,20 +981,16 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 32,
+   "execution_count": 33,
    "metadata": {
     "scrolled": true
    },
    "outputs": [],
    "source": [
-    "# create a folder for JIT cache\n",
-    "cache_dir = os.path.join(cwd, 'bidaf_cache')\n",
-    "# remove any stale cache files\n",
-    "if os.path.exists(cache_dir):\n",
-    "  shutil.rmtree(cache_dir)\n",
-    "os.makedirs(cache_dir, exist_ok=True)\n",
     "# use settings to enable JIT cache\n",
-    "settings = 'nuphar_cache_path:{}'.format(cache_dir)\n",
+    "bidaf_cache_dir = os.path.join(bidaf_dir, 'cache')\n",
+    "create_cache_dir(bidaf_cache_dir)\n",
+    "settings = 'nuphar_cache_path:{}'.format(bidaf_cache_dir)\n",
     "onnxruntime.capi._pybind_state.set_nuphar_settings(settings)\n",
     "sess = onnxruntime.InferenceSession(bidaf_converted)"
    ]
@@ -957,7 +1004,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 33,
+   "execution_count": 34,
    "metadata": {},
    "outputs": [
     {
@@ -966,17 +1013,16 @@
        "['jit.so']"
       ]
      },
-     "execution_count": 33,
+     "execution_count": 34,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "cache_versioned_dir = os.path.join(cache_dir, os.listdir(cache_dir)[0])\n",
+    "bidaf_cache_versioned_dir = os.path.join(bidaf_cache_dir, os.listdir(bidaf_cache_dir)[0])\n",
     "# use onnxruntime.nuphar.create_shared module to create dll\n",
-    "onnxruntime_dir = os.path.split(os.path.abspath(onnxruntime.__file__))[0]\n",
-    "subprocess.run([sys.executable, '-m', 'onnxruntime.nuphar.create_shared', '--input_dir', cache_versioned_dir], check=True)\n",
-    "os.listdir(cache_versioned_dir)"
+    "subprocess.run([sys.executable, '-m', 'onnxruntime.nuphar.create_shared', '--input_dir', bidaf_cache_versioned_dir], check=True)\n",
+    "os.listdir(bidaf_cache_versioned_dir)"
    ]
   },
   {
@@ -988,28 +1034,95 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 34,
+   "execution_count": 35,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "AOT speed-up 764.62%\n",
-      "    Baseline: 4.721 s, Current: 0.546 s\n"
+      "AOT speed-up 694.06%\n",
+      "    Baseline: 5.749 s, Current: 0.724 s\n"
      ]
     }
    ],
    "source": [
     "start_aot = timer()\n",
     "# NOTE: Nuphar settings string is not sticky. It needs to be reset before creating InferenceSession\n",
-    "settings = 'nuphar_cache_path:{}'.format(cache_dir)\n",
+    "settings = 'nuphar_cache_path:{}'.format(bidaf_cache_dir)\n",
     "onnxruntime.capi._pybind_state.set_nuphar_settings(settings)\n",
     "sess = onnxruntime.InferenceSession(bidaf_converted)\n",
     "end_aot = timer()\n",
     "print_speedup('AOT', end_jit - start_jit, end_aot - start_aot)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Moreover, Nuphar AOT also supports:\n",
+    "* Generate JIT cache with AVX/AVX2/AVX-512 and build a AOT dll including support for all these CPUs, which makes deployment easier when targeting different CPUs in one package.\n",
+    "* Bake model checksum into AOT dll to validate model with given AOT dll."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 36,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Scan_LSTMx4_int8.onnx in avx2 speed-up 790.69%\n",
+      "    Baseline: 3.070 s, Current: 0.345 s\n",
+      "Scan_LSTMx4_int8.onnx in avx speed-up 361.68%\n",
+      "    Baseline: 3.070 s, Current: 0.665 s\n"
+     ]
+    }
+   ],
+   "source": [
+    "# create object files for different CPUs\n",
+    "cache_dir = os.path.join(os.getcwd(), 'lstm_cache')\n",
+    "model_name = lstm_quantized_model\n",
+    "model_checksum = md5(model_name)\n",
+    "repeats = lstm_repeats\n",
+    "feed = lstm_feed\n",
+    "time_baseline = end_lstm_baseline - start_lstm_baseline\n",
+    "multi_isa_so = 'avx_avx2_avx512.so'\n",
+    "\n",
+    "create_cache_dir(cache_dir)\n",
+    "settings = 'nuphar_cache_path:{}'.format(cache_dir)\n",
+    "for isa in ['avx512', 'avx2', 'avx']:\n",
+    "    settings_with_isa = settings + ', nuphar_codegen_target:' + isa\n",
+    "    onnxruntime.capi._pybind_state.set_nuphar_settings(settings_with_isa)\n",
+    "    sess = onnxruntime.InferenceSession(model_name)\n",
+    "    cache_versioned_dir = os.path.join(cache_dir, os.listdir(cache_dir)[0])\n",
+    "\n",
+    "# link object files to AOT dll\n",
+    "subprocess.run([sys.executable, '-m', 'onnxruntime.nuphar.create_shared', '--input_dir', cache_versioned_dir, '--input_model', model_name, '--output_name', multi_isa_so], check=True)\n",
+    "\n",
+    "# now load the model with AOT dll\n",
+    "# NOTE: when nuphar_codegen_target is not set, it defaults to current CPU ISA\n",
+    "settings = 'nuphar_cache_path:{}, nuphar_cache_so_name:{}, nuphar_cache_model_checksum:{}, nuphar_cache_force_no_jit:on'.format(cache_dir, multi_isa_so, model_checksum)\n",
+    "onnxruntime.capi._pybind_state.set_nuphar_settings(settings)\n",
+    "sess = onnxruntime.InferenceSession(model_name)\n",
+    "\n",
+    "# force to a different ISA which is a subset of current CPU\n",
+    "# NOTE: if an incompatible ISA is used, exception on invalid instructions would be thrown\n",
+    "for valid_isa in ['avx2', 'avx']:\n",
+    "    settings_with_isa = 'nuphar_cache_path:{}, nuphar_cache_so_name:{}, nuphar_cache_model_checksum:{}, nuphar_codegen_target:{}, nuphar_cache_force_no_jit:on'.format(cache_dir, multi_isa_so, model_checksum, valid_isa)\n",
+    "    onnxruntime.capi._pybind_state.set_nuphar_settings(settings_with_isa)\n",
+    "    sess = onnxruntime.InferenceSession(model_name)\n",
+    "\n",
+    "    start_nuphar = timer()\n",
+    "    for i in range(repeats):\n",
+    "        sess.run([], feed)\n",
+    "    end_nuphar = timer()\n",
+    "\n",
+    "    print_speedup('{} in {}'.format(model_name, valid_isa), time_baseline, end_nuphar - start_nuphar)"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1020,15 +1133,15 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 35,
+   "execution_count": 37,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Single thread perf w/o parallel schedule speed-up 1.05%\n",
-      "    Baseline: 1.542 s, Current: 1.526 s\n"
+      "Single thread perf w/o parallel schedule speed-up 1.49%\n",
+      "    Baseline: 1.540 s, Current: 1.518 s\n"
      ]
     }
    ],
@@ -1039,8 +1152,8 @@
     "\n",
     "sess = onnxruntime.InferenceSession(bidaf_converted)\n",
     "start_baseline = timer()\n",
-    "for i in range(repeats):\n",
-    "    output_baseline = sess_baseline.run([], feed)\n",
+    "for i in range(bidaf_repeats):\n",
+    "    sess_baseline.run([], bidaf_feed)\n",
     "end_baseline = timer()\n",
     "\n",
     "# use NUPHAR_PARALLEL_MIN_WORKLOADS=0 to turn off parallel schedule, using settings string\n",
@@ -1050,10 +1163,12 @@
     "sess = onnxruntime.InferenceSession(bidaf_converted)\n",
     "\n",
     "start = timer()\n",
-    "for i in range(repeats):\n",
-    "    output = sess_baseline.run([], feed)\n",
+    "for i in range(bidaf_repeats):\n",
+    "    sess_baseline.run([], bidaf_feed)\n",
     "end = timer()\n",
-    "print_speedup('Single thread perf w/o parallel schedule', end_baseline - start_baseline, end - start)"
+    "print_speedup('Single thread perf w/o parallel schedule', end_baseline - start_baseline, end - start)\n",
+    "\n",
+    "del os.environ['OMP_NUM_THREADS']"
    ]
   }
  ],

From cd6ff42442b43de90785bcf0d37df575273c4959 Mon Sep 17 00:00:00 2001
From: Ashwini Khade <askhade@microsoft.com>
Date: Mon, 23 Dec 2019 10:19:28 -0800
Subject: [PATCH 06/75] Add schema for new Qops (#2611)

* Add schema for new Qops

* adding shape inference + qlinearaveragepool

* plus review comments

* plus review comments

* updates per review comments

* plus review comments
---
 .../core/graph/contrib_ops/contrib_defs.cc    | 422 +++++++++++++++++-
 1 file changed, 421 insertions(+), 1 deletion(-)

diff --git a/onnxruntime/core/graph/contrib_ops/contrib_defs.cc b/onnxruntime/core/graph/contrib_ops/contrib_defs.cc
index 422c1aad31f94..ccfe61344b190 100644
--- a/onnxruntime/core/graph/contrib_ops/contrib_defs.cc
+++ b/onnxruntime/core/graph/contrib_ops/contrib_defs.cc
@@ -162,7 +162,6 @@ void convTransposeWithDynamicPadsShapeInference(InferenceContext& ctx) {
     return;
   }
 }
-
 }  // namespace ONNX_NAMESPACE
 
 namespace onnxruntime {
@@ -202,6 +201,126 @@ void NchwcGlobalPoolOpSchemaGenerator(OpSchema& schema) {
   });
 }
 
+void ValidateTypeAndShapeForScaleAndZP(ONNX_NAMESPACE::InferenceContext& ctx, int index, ::google::protobuf::int32 expectedType, bool isScalar, int expectedTensorSize = 0) {
+  if (ctx.getNumInputs() > static_cast<size_t>(index)) {
+    auto data_type = ctx.getInputType(index);
+    if (nullptr == data_type || data_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType ||
+        data_type->tensor_type().elem_type() != expectedType) {
+      fail_type_inference(
+          "Input data type does not match the expected data type. Current data type is ", data_type->tensor_type().elem_type());
+    }
+  }
+
+  if (hasInputShape(ctx, index)) {
+    ONNX_NAMESPACE::TensorShapeProto shape = ctx.getInputType(index)->tensor_type().shape();
+    if (isScalar) {
+      if (shape.dim_size() != 0) {
+        fail_type_inference("Scale and Zero-point must be a scalar");
+      }
+    } else {
+      if (shape.dim_size() != 1) {
+        fail_type_inference("Scale and Zero-point must be of rank 1");
+      }
+
+      if (shape.dim((int)0).has_dim_value() && shape.dim((int)0).dim_value() != expectedTensorSize) {
+        fail_type_inference(
+            "Scale and Zero-point must be of rank 1 and the number of elements should be equal to the number of rows of the corresponding input.");
+      }
+    }
+  }
+}
+
+std::function<void(OpSchema&)> QLinearMathDocGenerator(const char* name, const char* additionalDocumentation) {
+  return [=](OpSchema& schema) {
+    std::string doc = R"DOC(
+Performs element-wise binary {name} on 8 bit data types (with Numpy-style broadcasting support).
+
+{additionalDocumentation}
+)DOC";
+    ONNX_NAMESPACE::ReplaceAll(doc, "{name}", name);
+    ONNX_NAMESPACE::ReplaceAll(doc, "{additionalDocumentation}", additionalDocumentation);
+    schema.SetDoc(doc);
+    schema.Input(0, "A", "First operand.", "T");
+    schema.Input(
+        1,
+        "A_scale",
+        "Input A's scale. It's a scalar, which means a per-tensor/layer quantization.",
+        "tensor(float)");
+    schema.Input(
+        2,
+        "A_zero_point",
+        "Input A zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+        "T",
+        OpSchema::Optional);
+    schema.Input(3, "B", "Second operand.", "T");
+    schema.Input(
+        4,
+        "B_scale",
+        "Input B's scale. It's a scalar, which means a per-tensor/layer quantization.",
+        "tensor(float)");
+    schema.Input(
+        5,
+        "B_zero_point",
+        "Input B zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+        "T",
+        OpSchema::Optional);
+    schema.Input(
+        6,
+        "C_scale",
+        "Output scale. It's a scalar, which means a per-tensor/layer quantization.",
+        "tensor(float)");
+    schema.Input(
+        7,
+        "C_zero_point",
+        "Output zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+        "T",
+        OpSchema::Optional);
+    schema.Output(0, "C", "Result, has same element type as two inputs", "T");
+    schema.TypeConstraint("T", {"tensor(uint8)", "tensor(int8)"}, "Constrain input and output types to 8 bit signed and unsigned tensors.");
+    schema.TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
+      propagateElemTypeFromInputToOutput(ctx, 0, 0);
+
+      auto a_type = ctx.getInputType(0);
+      auto b_type = ctx.getInputType(3);
+
+      if (nullptr == a_type || nullptr == b_type ||
+          a_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType ||
+          b_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType) {
+        fail_type_inference("inputs are expected to have tensor type.");
+      }
+
+      // validate scale and zero points
+      ValidateTypeAndShapeForScaleAndZP(ctx, 1, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+      ValidateTypeAndShapeForScaleAndZP(ctx, 2, a_type->tensor_type().elem_type(), true);
+      ValidateTypeAndShapeForScaleAndZP(ctx, 4, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+      ValidateTypeAndShapeForScaleAndZP(ctx, 5, b_type->tensor_type().elem_type(), true);
+      ValidateTypeAndShapeForScaleAndZP(ctx, 6, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+      ValidateTypeAndShapeForScaleAndZP(ctx, 7, a_type->tensor_type().elem_type(), true);
+
+      if (hasInputShape(ctx, 0) && hasInputShape(ctx, 3))
+        bidirectionalBroadcastShapeInference(
+            ctx.getInputType(0)->tensor_type().shape(),
+            ctx.getInputType(3)->tensor_type().shape(),
+            *ctx.getOutputType(0)->mutable_tensor_type()->mutable_shape());
+    });
+  };
+}
+
+const char* contrib_ops_pads_doc =
+    "Padding for the beginning and ending along each spatial axis, it can take any value greater "
+    "than or equal to 0. The value represent the number of pixels added to the beginning "
+    "and end part of the corresponding axis. `pads` format should be as follow "
+    "[x1_begin, x2_begin...x1_end, x2_end,...], where xi_begin the number of pixels "
+    "added at the beginning of axis `i` and xi_end, the number of pixels added at "
+    "the end of axis `i`. This attribute cannot be used simultaneously with "
+    "auto_pad attribute. If not present, the padding defaults to 0 along start and end of each spatial axis.";
+const char* contrib_ops_auto_pad_doc =
+    "auto_pad must be either NOTSET, SAME_UPPER, SAME_LOWER or VALID. Where "
+    "default value is NOTSET, which means explicit padding is used. "
+    "SAME_UPPER or SAME_LOWER mean pad the input so that the output spatial size match the input."
+    "In case of odd number add the extra padding at the end for SAME_UPPER and at the "
+    "beginning for SAME_LOWER. VALID mean no padding.";
+
 void RegisterNchwcSchemas() {
   ONNX_CONTRIB_OPERATOR_SCHEMA(ReorderInput)
       .SetDomain(kMSNchwcDomain)
@@ -1541,6 +1660,307 @@ with the exception that numpy default keepdims to False instead of True.)DOC")
           "Keep the reduced dimension or not, default 1 mean keep reduced dimension.",
           AttributeProto::INT);
 
+  ONNX_CONTRIB_OPERATOR_SCHEMA(QLinearAdd)
+      .SetDomain(kMSDomain)
+      .SinceVersion(1)
+      .FillUsing(QLinearMathDocGenerator("addition",
+                                         "C = (A_scale * (A - A_zero_point) + B_scale * (B - B_zero_point))/C_scale + C_zero_point"));
+
+  ONNX_CONTRIB_OPERATOR_SCHEMA(QLinearMul)
+      .SetDomain(kMSDomain)
+      .SinceVersion(1)
+      .FillUsing(QLinearMathDocGenerator("multiplication",
+                                         "C = ((A - A_zero_point) * (B - B_zero_point)) * (A_scale * B_scale)/C_scale + C_zero_point"));
+
+  ONNX_CONTRIB_OPERATOR_SCHEMA(QLinearReduceMean)
+      .SetDomain(kMSDomain)
+      .SinceVersion(1)
+      .SetDoc(R"DOC(
+Computes the mean of the low-precision input tensor's element along the provided axes.
+The resulting tensor has the same rank as the input if keepdims equal 1. If keepdims equal 0,
+then the resulting tensor have the reduced dimension pruned. The above behavior is similar to numpy,
+with the exception that numpy default keepdims to False instead of True.
+Input and Output scales and zero points are used to requantize the output in a new range. 
+This helps to improve accuracy as after ReduceMean operation the range of the output is expected to decrease.
+
+```
+"Output = Dequantize(Input) -> ReduceMean on fp32 data -> Quantize(output)",
+
+```
+)DOC")
+      .Input(0, "data", "An input tensor.", "T")
+      .Input(
+          1,
+          "data_scale",
+          "Input scale. It's a scalar, which means a per-tensor/layer quantization.",
+          "tensor(float)")
+      .Input(
+          2,
+          "data_zero_point",
+          "Input zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+          "T",
+          OpSchema::Optional)
+      .Input(
+          3,
+          "reduced_scale",
+          "Output scale. It's a scalar, which means a per-tensor/layer quantization.",
+          "tensor(float)")
+      .Input(
+          4,
+          "reduced_zero_point",
+          "Output zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+          "T",
+          OpSchema::Optional)
+      .Output(0, "reduced", "Reduced output tensor.", "T")
+      .TypeConstraint("T", {"tensor(uint8)", "tensor(int8)"}, "Constrain input types to 8 bit signed and unsigned tensors.")
+      .Attr(
+          "axes",
+          "A list of integers, along which to reduce. The default is to reduce over all the dimensions of the input tensor.",
+          AttributeProto::INTS)
+      .Attr(
+          "keepdims",
+          "Keep the reduced dimension or not, default 1 mean keep reduced dimension.",
+          AttributeProto::INT)
+      .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
+        propagateElemTypeFromInputToOutput(ctx, 0, 0);
+
+        if (!hasNInputShapes(ctx, 1)) {
+          return;
+        }
+
+        auto data_type = ctx.getInputType(0);
+        if (nullptr == data_type || data_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType) {
+          fail_type_inference("inputs are expected to have tensor type.");
+        }
+
+        // validate scale and zero points
+        ValidateTypeAndShapeForScaleAndZP(ctx, 1, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 2, data_type->tensor_type().elem_type(), true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 3, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 4, data_type->tensor_type().elem_type(), true);
+
+        int64_t keep_dims = 1;
+        auto attr_proto = ctx.getAttribute("keepdims");
+        if (attr_proto) {
+          keep_dims = attr_proto->i();
+        }
+
+        auto& input_shape = ctx.getInputType(0)->tensor_type().shape();
+        int64_t input_ndim = input_shape.dim_size();
+        auto output_shape =
+            ctx.getOutputType(0)->mutable_tensor_type()->mutable_shape();
+        std::vector<int64_t> axes;
+        auto axes_proto = ctx.getAttribute("axes");
+        if (axes_proto)
+          axes.assign(axes_proto->ints().begin(), axes_proto->ints().end());
+
+        for (size_t i = 0; i < axes.size(); ++i) {
+          if (axes[i] < -input_ndim || axes[i] >= input_ndim) {
+            fail_shape_inference(
+                "axis must be in [-rank, rank-1]. input rank was ", input_ndim);
+          }
+          if (axes[i] < 0)
+            axes[i] += input_ndim;
+        }
+        // do we need to handle negative axis?
+        for (int i = 0; i < input_ndim; ++i) {
+          // axes empty means reduce all dim
+          if (!axes.empty() &&
+              std::find(axes.begin(), axes.end(), i) == axes.end()) {
+            auto dim = output_shape->add_dim();
+            dim->CopyFrom(input_shape.dim(i));
+          } else {
+            if (keep_dims == 1) {
+              auto dim = output_shape->add_dim();
+              dim->set_dim_value(1);
+            }
+          }
+        }
+      });
+
+  ONNX_CONTRIB_OPERATOR_SCHEMA(MulInteger)
+      .SetDomain(kMSDomain)
+      .SinceVersion(1)
+      .SetDoc(R"DOC(Performs element-wise binary quantized multiplication (with Numpy-style broadcasting support).
+"This operator supports **multidirectional (i.e., Numpy-style) broadcasting**"
+The output of this op is the int32 accumulated result of the mul operation
+
+```
+C (int32) = (A - A_zero_point) * (B - B_zero_point)
+```
+
+)DOC")
+      .Input(0, "A", "First operand.", "T")
+      .Input(
+          1,
+          "A_zero_point",
+          "Input A zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+          "T",
+          OpSchema::Optional)
+      .Input(2, "B", "Second operand.", "T")
+      .Input(
+          3,
+          "B_zero_point",
+          "Input B zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+          "T",
+          OpSchema::Optional)
+      .Output(0, "C", "Constrain output to 32 bit tensor", "T1")
+      .TypeConstraint("T", {"tensor(uint8)", "tensor(int8)"}, "Constrain input types to 8 bit signed and unsigned tensors.")
+      .TypeConstraint("T1", {"tensor(int32)"}, "Constrain output types to 32 bit tensors.")
+      .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
+        auto c_type = ctx.getOutputType(0);
+        c_type->mutable_tensor_type()->set_elem_type(
+            ONNX_NAMESPACE::TensorProto::INT32);
+
+        auto a_type = ctx.getInputType(0);
+        auto b_type = ctx.getInputType(3);
+        if (nullptr == a_type || nullptr == b_type ||
+            a_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType ||
+            b_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType) {
+          fail_type_inference("inputs are expected to have tensor type.");
+        }
+
+        ValidateTypeAndShapeForScaleAndZP(ctx, 1, a_type->tensor_type().elem_type(), true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 3, b_type->tensor_type().elem_type(), true);
+
+        if (hasInputShape(ctx, 0) && hasInputShape(ctx, 2)) {
+          bidirectionalBroadcastShapeInference(
+              ctx.getInputType(0)->tensor_type().shape(),
+              ctx.getInputType(2)->tensor_type().shape(),
+              *ctx.getOutputType(0)->mutable_tensor_type()->mutable_shape());
+        }
+      });
+
+  const char* QLinearAveragePoolDoc_ver1 = R"DOC(
+ QLinearAveragePool consumes an input tensor X and applies average pooling across
+ the tensor according to kernel sizes, stride sizes, and pad lengths.
+ average pooling consisting of computing the average on all values of a
+ subset of the input tensor according to the kernel size and downsampling the
+ data into the output tensor Y for further processing. The output spatial shape will be following:
+ ```
+ output_spatial_shape[i] = floor((input_spatial_shape[i] + pad_shape[i] - kernel_spatial_shape[i]) / strides_spatial_shape[i] + 1)
+ ```
+ or
+ ```
+ output_spatial_shape[i] = ceil((input_spatial_shape[i] + pad_shape[i] - kernel_spatial_shape[i]) / strides_spatial_shape[i] + 1)
+ ```
+ if ceil_mode is enabled
+
+ ```
+ * pad_shape[i] is sum of pads along axis i
+ ```
+
+ `auto_pad` is a DEPRECATED attribute. If you are using them currently, the output spatial shape will be following:
+ ```
+ VALID: output_spatial_shape[i] = ceil((input_spatial_shape[i] - kernel_spatial_shape[i] + 1) / strides_spatial_shape[i])
+ SAME_UPPER or SAME_LOWER: output_spatial_shape[i] = ceil(input_spatial_shape[i] / strides_spatial_shape[i])
+ ```
+ And pad shape will be following if `SAME_UPPER` or `SAME_LOWER`:
+ ```
+ pad_shape[i] = (output_spatial_shape[i] - 1) * strides_spatial_shape[i] + kernel_spatial_shape[i] - input_spatial_shape[i]
+ ```
+ 
+The output of each pooling window is divided by the number of elements (exclude pad when attribute count_include_pad is zero).
+
+Input and output scales and zero points are used to convert the output to a new quantization range.
+Output = Dequantize(Input) -> AveragePool on fp32 data -> Quantize(output)
+)DOC";
+
+  ONNX_CONTRIB_OPERATOR_SCHEMA(QLinearAveragePool)
+      .SetDomain(kMSDomain)
+      .SinceVersion(1)
+      .SetDoc(QLinearAveragePoolDoc_ver1)
+      .Attr(
+          "count_include_pad",
+          "Whether include pad pixels when calculating values for the edges. Default is 0, doesn't count include pad.",
+          AttributeProto::INT,
+          static_cast<int64_t>(0))
+      .Attr(
+          "kernel_shape",
+          "The size of the kernel along each axis.",
+          AttributeProto::INTS)
+      .Attr(
+          "strides",
+          "Stride along each spatial axis. If not present, the stride defaults to 1 along each spatial axis.",
+          AttributeProto::INTS,
+          OPTIONAL)
+      .Attr(
+          "auto_pad",
+          contrib_ops_auto_pad_doc,
+          AttributeProto::STRING,
+          std::string("NOTSET"))
+      .Attr("pads", contrib_ops_pads_doc, AttributeProto::INTS, OPTIONAL)
+      .Attr(
+          "ceil_mode",
+          "Whether to use ceil or floor (default) to compute the output shape.",
+          AttributeProto::INT,
+          static_cast<int64_t>(0))
+      .Input(
+          0,
+          "X",
+          "Input data tensor from the previous operator; "
+          "dimensions for image case are (N x C x H x W), "
+          "where N is the batch size, C is the number of "
+          "channels, and H and W are the height and the "
+          "width of the data. For non image case, the "
+          "dimensions are in the form of "
+          "(N x C x D1 x D2 ... Dn), where N is the batch "
+          "size. Optionally, if dimension denotation is "
+          "in effect, the operation expects the input "
+          "data tensor to arrive with the dimension denotation "
+          "of [DATA_BATCH, DATA_CHANNEL, DATA_FEATURE, DATA_FEATURE ...].",
+          "T")
+      .Input(
+          1,
+          "x_scale",
+          "Input scale. It's a scalar, which means a per-tensor/layer quantization.",
+          "tensor(float)")
+      .Input(
+          2,
+          "x_zero_point",
+          "Input zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+          "T",
+          OpSchema::Optional)
+      .Input(
+          3,
+          "y_scale",
+          "Output scale. It's a scalar, which means a per-tensor/layer quantization.",
+          "tensor(float)")
+      .Input(
+          4,
+          "y_zero_point",
+          "Output zero point. Default value is 0 if it's not specified. It's a scalar, which means a per-tensor/layer quantization.",
+          "T",
+          OpSchema::Optional)
+      .Output(
+          0,
+          "Y",
+          "Output data tensor from average or max pooling across "
+          "the input tensor. Dimensions will vary based "
+          "on various kernel, stride, and pad sizes. Floor value of "
+          "the dimension is used",
+          "T")
+      .TypeConstraint(
+          "T",
+          {"tensor(uint8)", "tensor(int8)"},
+          "Constrain input and output types to 8 bit tensors.")
+      .TypeAndShapeInferenceFunction([](ONNX_NAMESPACE::InferenceContext& ctx) {
+        ONNX_NAMESPACE::propagateElemTypeFromInputToOutput(ctx, 0, 0);
+
+        auto data_type = ctx.getInputType(0);
+        if (nullptr == data_type || data_type->value_case() != ONNX_NAMESPACE::TypeProto::kTensorType) {
+          fail_type_inference("inputs are expected to have tensor type.");
+        }
+
+        // validate scale and zero points
+        ValidateTypeAndShapeForScaleAndZP(ctx, 1, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 2, data_type->tensor_type().elem_type(), true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 3, ONNX_NAMESPACE::TensorProto::FLOAT, true);
+        ValidateTypeAndShapeForScaleAndZP(ctx, 4, data_type->tensor_type().elem_type(), true);
+
+        ONNX_NAMESPACE::convPoolShapeInference(ctx, false, true, 0, 5);
+      });
+
   ONNX_CONTRIB_OPERATOR_SCHEMA(MurmurHash3)
       .SetDomain(kMSDomain)
       .SinceVersion(1)

From abda4780839119668893a652caf78d65542d18cb Mon Sep 17 00:00:00 2001
From: Andrey <andrey.lompart@gmail.com>
Date: Tue, 24 Dec 2019 04:52:09 +0800
Subject: [PATCH 07/75] [server] Add supposed for model_name and model_version
 as cli parameter (#2708)

---
 onnxruntime/server/main.cc                             |  6 ++++--
 onnxruntime/server/server_configuration.h              |  4 ++++
 .../server/unit_tests/server_configuration_test.cc     | 10 ++++++++--
 3 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/onnxruntime/server/main.cc b/onnxruntime/server/main.cc
index 9c0e98ec44719..bf039d8a4ae90 100644
--- a/onnxruntime/server/main.cc
+++ b/onnxruntime/server/main.cc
@@ -60,10 +60,12 @@ int main(int argc, char* argv[]) {
 
   const auto env = std::make_shared<server::ServerEnvironment>(config.logging_level, spdlog::sinks_init_list{std::make_shared<spdlog::sinks::stdout_sink_mt>(), std::make_shared<spdlog::sinks::syslog_sink_mt>()});
   auto logger = env->GetAppLogger();
-  logger->info("Model path: {}", config.model_path);
+  logger->info("Model path: {}, ", config.model_path);
+  logger->info("Model name: {}", config.model_name);
+  logger->info("Model version: {}", config.model_version);
 
   try {
-    env->InitializeModel(config.model_path, "default", "1");
+    env->InitializeModel(config.model_path, config.model_name, config.model_version);
     logger->debug("Initialize Model Successfully!");
   } catch (const Ort::Exception& ex) {
     logger->critical("Initialize Model Failed: {} ---- Error: [{}]", ex.GetOrtErrorCode(), ex.what());
diff --git a/onnxruntime/server/server_configuration.h b/onnxruntime/server/server_configuration.h
index 93d6983d33e51..3fb075a4776f4 100644
--- a/onnxruntime/server/server_configuration.h
+++ b/onnxruntime/server/server_configuration.h
@@ -39,6 +39,8 @@ class ServerConfiguration {
  public:
   const std::string full_desc = "ONNX Server: host an ONNX model with ONNX Runtime";
   std::string model_path;
+  std::string model_name = "default";
+  std::string model_version = "1";
   std::string address = "0.0.0.0";
   unsigned short http_port = 8001;
   unsigned short grpc_port = 50051;
@@ -49,6 +51,8 @@ class ServerConfiguration {
     desc.add_options()("help,h", "Shows a help message and exits");
     desc.add_options()("log_level", po::value(&log_level_str)->default_value(log_level_str), "Logging level. Allowed options (case sensitive): verbose, info, warning, error, fatal");
     desc.add_options()("model_path", po::value(&model_path)->required(), "Path to ONNX model");
+    desc.add_options()("model_name", po::value(&model_name)->default_value(model_name), "ONNX model name");
+    desc.add_options()("model_version", po::value(&model_version)->default_value(model_version), "ONNX model version");
     desc.add_options()("address", po::value(&address)->default_value(address), "The base HTTP address");
     desc.add_options()("http_port", po::value(&http_port)->default_value(http_port), "HTTP port to listen to requests");
     desc.add_options()("num_http_threads", po::value(&num_http_threads)->default_value(num_http_threads), "Number of http threads");
diff --git a/onnxruntime/test/server/unit_tests/server_configuration_test.cc b/onnxruntime/test/server/unit_tests/server_configuration_test.cc
index 00e772d902274..46e2b06b32fbe 100644
--- a/onnxruntime/test/server/unit_tests/server_configuration_test.cc
+++ b/onnxruntime/test/server/unit_tests/server_configuration_test.cc
@@ -14,15 +14,19 @@ TEST(ConfigParsingTests, AllArgs) {
   char* test_argv[] = {
       const_cast<char*>("/path/to/binary"),
       const_cast<char*>("--model_path"), const_cast<char*>("testdata/mul_1.onnx"),
+      const_cast<char*>("--model_name"), const_cast<char*>("mul_1"),
+      const_cast<char*>("--model_version"), const_cast<char*>("1"),
       const_cast<char*>("--address"), const_cast<char*>("4.4.4.4"),
       const_cast<char*>("--http_port"), const_cast<char*>("80"),
       const_cast<char*>("--num_http_threads"), const_cast<char*>("1"),
       const_cast<char*>("--log_level"), const_cast<char*>("info")};
 
   onnxruntime::server::ServerConfiguration config{};
-  Result res = config.ParseInput(11, test_argv);
+  Result res = config.ParseInput(15, test_argv);
   EXPECT_EQ(res, Result::ContinueSuccess);
   EXPECT_EQ(config.model_path, "testdata/mul_1.onnx");
+  EXPECT_EQ(config.model_name, "mul_1");
+  EXPECT_EQ(config.model_version, "1");
   EXPECT_EQ(config.address, "4.4.4.4");
   EXPECT_EQ(config.http_port, 80);
   EXPECT_EQ(config.num_http_threads, 1);
@@ -32,13 +36,15 @@ TEST(ConfigParsingTests, AllArgs) {
 TEST(ConfigParsingTests, Defaults) {
   char* test_argv[] = {
       const_cast<char*>("/path/to/binary"),
-      const_cast<char*>("--model"), const_cast<char*>("testdata/mul_1.onnx"),
+      const_cast<char*>("--model_path"), const_cast<char*>("testdata/mul_1.onnx"),
       const_cast<char*>("--num_http_threads"), const_cast<char*>("3")};
 
   onnxruntime::server::ServerConfiguration config{};
   Result res = config.ParseInput(5, test_argv);
   EXPECT_EQ(res, Result::ContinueSuccess);
   EXPECT_EQ(config.model_path, "testdata/mul_1.onnx");
+  EXPECT_EQ(config.model_name, "default");
+  EXPECT_EQ(config.model_version, "1");
   EXPECT_EQ(config.address, "0.0.0.0");
   EXPECT_EQ(config.http_port, 8001);
   EXPECT_EQ(config.num_http_threads, 3);

From 71ce0b8e0a35a2f762a397fbb4d54f02dc8d64de Mon Sep 17 00:00:00 2001
From: George Wu <George.Wu@microsoft.com>
Date: Mon, 23 Dec 2019 17:06:55 -0800
Subject: [PATCH 08/75] remove 64bit warning message from python validation.
 (#2727)

---
 onnxruntime/python/onnxruntime_validation.py | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/onnxruntime/python/onnxruntime_validation.py b/onnxruntime/python/onnxruntime_validation.py
index 0cb88d3505014..73bd352c32556 100644
--- a/onnxruntime/python/onnxruntime_validation.py
+++ b/onnxruntime/python/onnxruntime_validation.py
@@ -13,7 +13,6 @@ def check_distro_info():
     __my_distro__ = ''
     __my_distro_ver__ = ''
     __my_system__ = platform.system().lower()
-    __my_arch__ = platform.architecture()[0].lower()
 
     __OS_RELEASE_FILE__ = '/etc/os-release'
     __LSB_RELEASE_FILE__ = '/etc/lsb-release'
@@ -25,8 +24,6 @@ def check_distro_info():
         if __my_distro_ver__ != '10':
             warnings.warn('Unsupported Windows version (%s). ONNX Runtime supports Windows 10 and above, only.' % __my_distro_ver__)
     elif __my_system__ == 'linux':
-        if __my_arch__ != '64bit':
-            warnings.warn('Unsupported architecture (%s). ONNX Runtime supports 64bit architecture, only.' % __my_arch__)
         ''' Although the 'platform' python module for getting Distro information works well on standard OS images running on real
         hardware, it is not acurate when running on Azure VMs, Git Bash, Cygwin, etc. The returned values for release and version
         are unpredictable for virtualized or emulated environments.

From ebf23744eb1e587e12e3459b1891a89a840fc687 Mon Sep 17 00:00:00 2001
From: Tracy Sharpe <42477615+tracysh@users.noreply.github.com>
Date: Tue, 24 Dec 2019 19:03:05 -0800
Subject: [PATCH 09/75] MLAS: ARM64 build fix (#2734)

fix bad usage of vreinterpret to cast vector element types
---
 onnxruntime/core/mlas/lib/quantize.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/onnxruntime/core/mlas/lib/quantize.cpp b/onnxruntime/core/mlas/lib/quantize.cpp
index 2a04b14d25083..c81897d6f9044 100644
--- a/onnxruntime/core/mlas/lib/quantize.cpp
+++ b/onnxruntime/core/mlas/lib/quantize.cpp
@@ -176,7 +176,7 @@ MlasQuantizeLinearKernel(
             MinimumValueVector, MaximumValueVector, ZeroPointVector);
 
 #if defined(MLAS_NEON64_INTRINSICS)
-        vst1q_lane_u8((uint8_t*)Output + n, vreinterpretq_s32_u8(IntegerVector), 0);
+        vst1q_lane_u8((uint8_t*)Output + n, vreinterpretq_u8_s32(IntegerVector), 0);
 #else
         *((uint8_t*)Output + n) = (uint8_t)_mm_cvtsi128_si32(IntegerVector);
 #endif

From f142b683a129861cde6d73b120ee07a20df4773c Mon Sep 17 00:00:00 2001
From: Faith Xu <faxu@microsoft.com>
Date: Wed, 25 Dec 2019 15:58:56 -0800
Subject: [PATCH 10/75] Fix broken python docs links (#2740)

---
 samples/README.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/samples/README.md b/samples/README.md
index 7b399953ff83e..50cada3d13a56 100644
--- a/samples/README.md
+++ b/samples/README.md
@@ -18,8 +18,8 @@ For a list of available dockerfiles and published images to help with getting st
 * [Model Inferencing using NUPHAR Execution Provider](../docs/python/notebooks/onnxruntime-nuphar-tutorial.ipynb)
 
 **Inference with model conversion**
-* [SKL Pipeline: Train, Convert, and Inference](https://microsoft.github.io/onnxruntime/tutorial.html)
-* [Keras: Convert and Inference](https://microsoft.github.io/onnxruntime/auto_examples/plot_dl_keras.html#sphx-glr-auto-examples-plot-dl-keras-py)
+* [SKL Pipeline: Train, Convert, and Inference](https://microsoft.github.io/onnxruntime/python/tutorial.html)
+* [Keras: Convert and Inference](https://microsoft.github.io/onnxruntime/python/auto_examples/plot_dl_keras.html#sphx-glr-auto-examples-plot-dl-keras-py)
 
 **Inference and deploy through AzureML**
 * Inferencing on CPU using [ONNX Model Zoo](https://github.com/onnx/models) models: 
@@ -43,7 +43,7 @@ For a list of available dockerfiles and published images to help with getting st
 
 **Other**
 * [Running ONNX model tests](./docs/Model_Test.md)
-* [Common Errors with explanations](https://microsoft.github.io/onnxruntime/auto_examples/plot_common_errors.html#sphx-glr-auto-examples-plot-common-errors-py)
+* [Common Errors with explanations](https://microsoft.github.io/onnxruntime/python/auto_examples/plot_common_errors.html#sphx-glr-auto-examples-plot-common-errors-py)
 
 ## C#
 * [Inferencing Tutorial](../docs/CSharp_API.md#getting-started)

From 9013fe6340574264d3cf33a430da0adc7a708c0b Mon Sep 17 00:00:00 2001
From: Andrey <andrey.lompart@gmail.com>
Date: Thu, 26 Dec 2019 12:10:39 +0800
Subject: [PATCH 11/75] Fix build on Mac OS (#2731)

mac os ld doesn't support --while-archive, correct option is -all_load
---
 cmake/onnxruntime_server.cmake | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/cmake/onnxruntime_server.cmake b/cmake/onnxruntime_server.cmake
index 86296105bef4e..d99ee7f48dd64 100644
--- a/cmake/onnxruntime_server.cmake
+++ b/cmake/onnxruntime_server.cmake
@@ -92,7 +92,11 @@ add_custom_command(
 
 add_library(server_grpc_proto ${grpc_srcs})
 target_include_directories(server_grpc_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_INCLUDE_DIRECTORIES> "${CMAKE_CURRENT_BINARY_DIR}" ${CMAKE_CURRENT_BINARY_DIR}/onnx PRIVATE)
-set(grpc_reflection -Wl,--whole-archive grpc++_reflection -Wl,--no-whole-archive)
+if(APPLE)
+  set(grpc_reflection -Wl,-all_load grpc++_reflection -Wl,-noall_load)
+else()
+  set(grpc_reflection -Wl,--whole-archive grpc++_reflection -Wl,--no-whole-archive)
+endif()
 set(grpc_static_libs grpc++ grpcpp_channelz)
 target_link_libraries(server_grpc_proto ${grpc_static_libs})
 add_dependencies(server_grpc_proto server_proto)

From 6e4ec9b049d43bdb11c2ee45f622ea1e19ccacc6 Mon Sep 17 00:00:00 2001
From: zhanyi-ms <zhanyi@microsoft.com>
Date: Fri, 27 Dec 2019 21:40:01 +0800
Subject: [PATCH 12/75] fix ngraph wheel (#2737)

* fix ngraph wheel

1.1.0 onnxruntime_ngraph wheel doesn't work

* remove libdnnl.so in nGraph Libs

* make it easy to compare
---
 setup.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index b53a6519acdd5..1347ccf318a91 100644
--- a/setup.py
+++ b/setup.py
@@ -119,7 +119,7 @@ def run(self):
 if platform.system() == 'Linux':
   libs = ['onnxruntime_pybind11_state.so', 'libdnnl.so.1', 'libmklml_intel.so', 'libiomp5.so', 'mimalloc.so']
   # nGraph Libs
-  libs.extend(['libngraph.so', 'libcodegen.so', 'libcpu_backend.so', 'libdnnl.so', 'libtbb_debug.so', 'libtbb_debug.so.2', 'libtbb.so', 'libtbb.so.2'])
+  libs.extend(['libngraph.so', 'libcodegen.so', 'libcpu_backend.so', 'libmkldnn.so', 'libtbb_debug.so', 'libtbb_debug.so.2', 'libtbb.so', 'libtbb.so.2'])
   # Nuphar Libs
   libs.extend(['libtvm.so.0.5.1'])
   # Openvino Libs

From c7a9c6b488c43990a532350cc788478e9585395d Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 27 Dec 2019 11:21:23 -0800
Subject: [PATCH 13/75] Split onnxruntime server to a separated folder (#2744)

---
 .gitmodules                                   |    9 +-
 cmake/external/grpc                           |    1 -
 cmake/onnxruntime_server.cmake                |  225 ----
 cmake/onnxruntime_unittests.cmake             |   81 --
 .../test/server/unit_tests/converter_tests.cc | 1113 -----------------
 server/CMakeLists.txt                         |  258 ++++
 server/ci/run.sh                              |   11 +
 server/cmake/FindProtobufWithTargets.cmake    |  203 +++
 server/cmake/FindgRPC.cmake                   |  353 ++++++
 {onnxruntime/server => server}/converter.cc   |   33 +-
 {onnxruntime/server => server}/converter.h    |    6 +-
 .../server => server}/core/request_id.cc      |    3 +
 .../server => server}/core/request_id.h       |    3 +
 {onnxruntime/server => server}/environment.cc |    2 +-
 {onnxruntime/server => server}/environment.h  |    2 +-
 {onnxruntime/server => server}/executor.cc    |    6 -
 {onnxruntime/server => server}/executor.h     |    2 +-
 {cmake => server}/external/spdlog             |    0
 {cmake => server}/get_boost.cmake             |    0
 .../server => server}/grpc/grpc_app.cc        |    3 +
 .../server => server}/grpc/grpc_app.h         |    3 +
 .../grpc/prediction_service_impl.cc           |    3 +
 .../grpc/prediction_service_impl.h            |    3 +
 .../server => server}/http/core/context.h     |    0
 .../http/core/http_server.cc                  |    0
 .../server => server}/http/core/http_server.h |    0
 .../server => server}/http/core/listener.cc   |    0
 .../server => server}/http/core/listener.h    |    0
 .../server => server}/http/core/routes.cc     |    0
 .../server => server}/http/core/routes.h      |    0
 .../server => server}/http/core/session.cc    |    0
 .../server => server}/http/core/session.h     |    0
 .../server => server}/http/core/util.cc       |    0
 .../server => server}/http/core/util.h        |    0
 .../server => server}/http/json_handling.cc   |    0
 .../server => server}/http/json_handling.h    |    0
 .../http/predict_request_handler.cc           |    0
 .../http/predict_request_handler.h            |    0
 {onnxruntime/server => server}/http/util.cc   |    0
 {onnxruntime/server => server}/http/util.h    |    0
 {onnxruntime/server => server}/main.cc        |    0
 .../server => server}/protobuf/onnx-ml.proto  |    0
 .../server => server}/protobuf/predict.proto  |    0
 .../protobuf/prediction_service.proto         |    0
 server/protobuf_function.cmake                |  158 +++
 .../serializing/mem_buffer.h                  |    8 +-
 .../serializing/tensorprotoutils.cc           |  138 +-
 .../serializing/tensorprotoutils.h            |    6 +-
 .../server => server}/server_configuration.h  |    2 +-
 .../test}/integration_tests/README.MD         |    0
 .../test}/integration_tests/function_tests.py |    0
 .../integration_tests/model_zoo_data_prep.py  |    0
 .../integration_tests/model_zoo_tests.py      |    0
 .../test}/integration_tests/test_main.py      |    0
 .../test}/integration_tests/test_util.py      |    0
 server/test/testdata/mul_1.onnx               |  Bin 0 -> 130 bytes
 .../server/mnist_test_data_set_0_input.json   |    0
 .../server/mnist_test_data_set_0_input.pb     |  Bin
 .../server/mnist_test_data_set_0_output.json  |    0
 .../server/mnist_test_data_set_0_output.pb    |    0
 .../test/testdata/server/request_0.pb         |  Bin
 .../test/testdata/server/response_0.pb        |    0
 .../test}/unit_tests/executor_test.cc         |    4 +-
 .../external/server_context_test_spouse.h     |    0
 .../test}/unit_tests/http_routes_tests.cc     |    2 +-
 .../test}/unit_tests/json_handling_tests.cc   |    6 +-
 .../prediction_service_impl_test.cc           |    7 +-
 .../unit_tests/server_configuration_test.cc   |    2 +-
 .../test}/unit_tests/test_main.cc             |    5 +-
 .../unit_tests/test_server_environment.cc     |   11 +-
 .../unit_tests/test_server_environment.h      |    7 +-
 .../test}/unit_tests/util_tests.cc            |    4 +-
 {onnxruntime/server => server}/util.cc        |   23 +-
 {onnxruntime/server => server}/util.h         |    6 +-
 tools/ci_build/build.py                       |   86 +-
 .../linux-ort-srv-ci-pipeline.yml             |   13 +-
 .../linux/docker/Dockerfile.ubuntu_server     |    2 +-
 .../docker/scripts/install_server_deps.sh     |   67 +-
 .../ci_build/github/linux/server_run_build.sh |   29 -
 .../github/linux/server_run_dockerbuild.sh    |   81 --
 80 files changed, 1188 insertions(+), 1802 deletions(-)
 delete mode 160000 cmake/external/grpc
 delete mode 100644 cmake/onnxruntime_server.cmake
 delete mode 100644 onnxruntime/test/server/unit_tests/converter_tests.cc
 create mode 100755 server/CMakeLists.txt
 create mode 100755 server/ci/run.sh
 create mode 100644 server/cmake/FindProtobufWithTargets.cmake
 create mode 100644 server/cmake/FindgRPC.cmake
 rename {onnxruntime/server => server}/converter.cc (88%)
 rename {onnxruntime/server => server}/converter.h (88%)
 rename {onnxruntime/server => server}/core/request_id.cc (87%)
 rename {onnxruntime/server => server}/core/request_id.h (75%)
 rename {onnxruntime/server => server}/environment.cc (99%)
 rename {onnxruntime/server => server}/environment.h (97%)
 rename {onnxruntime/server => server}/executor.cc (96%)
 rename {onnxruntime/server => server}/executor.h (97%)
 rename {cmake => server}/external/spdlog (100%)
 rename {cmake => server}/get_boost.cmake (100%)
 rename {onnxruntime/server => server}/grpc/grpc_app.cc (91%)
 rename {onnxruntime/server => server}/grpc/grpc_app.h (87%)
 rename {onnxruntime/server => server}/grpc/prediction_service_impl.cc (94%)
 rename {onnxruntime/server => server}/grpc/prediction_service_impl.h (89%)
 rename {onnxruntime/server => server}/http/core/context.h (100%)
 rename {onnxruntime/server => server}/http/core/http_server.cc (100%)
 rename {onnxruntime/server => server}/http/core/http_server.h (100%)
 rename {onnxruntime/server => server}/http/core/listener.cc (100%)
 rename {onnxruntime/server => server}/http/core/listener.h (100%)
 rename {onnxruntime/server => server}/http/core/routes.cc (100%)
 rename {onnxruntime/server => server}/http/core/routes.h (100%)
 rename {onnxruntime/server => server}/http/core/session.cc (100%)
 rename {onnxruntime/server => server}/http/core/session.h (100%)
 rename {onnxruntime/server => server}/http/core/util.cc (100%)
 rename {onnxruntime/server => server}/http/core/util.h (100%)
 rename {onnxruntime/server => server}/http/json_handling.cc (100%)
 rename {onnxruntime/server => server}/http/json_handling.h (100%)
 rename {onnxruntime/server => server}/http/predict_request_handler.cc (100%)
 rename {onnxruntime/server => server}/http/predict_request_handler.h (100%)
 rename {onnxruntime/server => server}/http/util.cc (100%)
 rename {onnxruntime/server => server}/http/util.h (100%)
 rename {onnxruntime/server => server}/main.cc (100%)
 rename {onnxruntime/server => server}/protobuf/onnx-ml.proto (100%)
 rename {onnxruntime/server => server}/protobuf/predict.proto (100%)
 rename {onnxruntime/server => server}/protobuf/prediction_service.proto (100%)
 create mode 100644 server/protobuf_function.cmake
 rename {onnxruntime/server => server}/serializing/mem_buffer.h (72%)
 rename {onnxruntime/server => server}/serializing/tensorprotoutils.cc (82%)
 rename {onnxruntime/server => server}/serializing/tensorprotoutils.h (91%)
 rename {onnxruntime/server => server}/server_configuration.h (99%)
 rename {onnxruntime/test/server => server/test}/integration_tests/README.MD (100%)
 rename {onnxruntime/test/server => server/test}/integration_tests/function_tests.py (100%)
 rename {onnxruntime/test/server => server/test}/integration_tests/model_zoo_data_prep.py (100%)
 rename {onnxruntime/test/server => server/test}/integration_tests/model_zoo_tests.py (100%)
 rename {onnxruntime/test/server => server/test}/integration_tests/test_main.py (100%)
 rename {onnxruntime/test/server => server/test}/integration_tests/test_util.py (100%)
 create mode 100644 server/test/testdata/mul_1.onnx
 rename {onnxruntime => server}/test/testdata/server/mnist_test_data_set_0_input.json (100%)
 rename {onnxruntime => server}/test/testdata/server/mnist_test_data_set_0_input.pb (100%)
 rename {onnxruntime => server}/test/testdata/server/mnist_test_data_set_0_output.json (100%)
 rename {onnxruntime => server}/test/testdata/server/mnist_test_data_set_0_output.pb (100%)
 rename {onnxruntime => server}/test/testdata/server/request_0.pb (100%)
 rename {onnxruntime => server}/test/testdata/server/response_0.pb (100%)
 rename {onnxruntime/test/server => server/test}/unit_tests/executor_test.cc (96%)
 rename {onnxruntime/test/server => server/test}/unit_tests/external/server_context_test_spouse.h (100%)
 rename {onnxruntime/test/server => server/test}/unit_tests/http_routes_tests.cc (99%)
 rename {onnxruntime/test/server => server/test}/unit_tests/json_handling_tests.cc (98%)
 rename {onnxruntime/test/server => server/test}/unit_tests/prediction_service_impl_test.cc (96%)
 rename {onnxruntime/test/server => server/test}/unit_tests/server_configuration_test.cc (98%)
 rename {onnxruntime/test/server => server/test}/unit_tests/test_main.cc (81%)
 rename {onnxruntime/test/server => server/test}/unit_tests/test_server_environment.cc (71%)
 rename {onnxruntime/test/server => server/test}/unit_tests/test_server_environment.h (71%)
 rename {onnxruntime/test/server => server/test}/unit_tests/util_tests.cc (98%)
 rename {onnxruntime/server => server}/util.cc (60%)
 rename {onnxruntime/server => server}/util.h (79%)
 delete mode 100755 tools/ci_build/github/linux/server_run_build.sh
 delete mode 100755 tools/ci_build/github/linux/server_run_dockerbuild.sh

diff --git a/.gitmodules b/.gitmodules
index accbde3a0c93d..19ec40f0ed5b5 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -25,15 +25,9 @@
 [submodule "cmake/external/eigen"]
 	path = cmake/external/eigen
 	url = https://github.com/eigenteam/eigen-git-mirror.git
-[submodule "cmake/external/grpc"]
-	path = cmake/external/grpc
-	url = https://github.com/grpc/grpc
 [submodule "cmake/external/DNNLibrary"]
 	path = cmake/external/DNNLibrary
 	url = https://github.com/JDAI-CV/DNNLibrary
-[submodule "cmake/external/spdlog"]
-	path = cmake/external/spdlog
-	url = https://github.com/gabime/spdlog.git
 [submodule "cmake/external/mimalloc"]
 	path = cmake/external/mimalloc
 	url = https://github.com/microsoft/mimalloc.git
@@ -49,3 +43,6 @@
 [submodule "cmake/external/json"]
 	path = cmake/external/json
 	url = https://github.com/nlohmann/json
+[submodule "server/external/spdlog"]
+	path = server/external/spdlog
+	url = https://github.com/gabime/spdlog.git
diff --git a/cmake/external/grpc b/cmake/external/grpc
deleted file mode 160000
index 02a2a458ac159..0000000000000
--- a/cmake/external/grpc
+++ /dev/null
@@ -1 +0,0 @@
-Subproject commit 02a2a458ac15912d7d87cc1171e811b0c5219ece
diff --git a/cmake/onnxruntime_server.cmake b/cmake/onnxruntime_server.cmake
deleted file mode 100644
index d99ee7f48dd64..0000000000000
--- a/cmake/onnxruntime_server.cmake
+++ /dev/null
@@ -1,225 +0,0 @@
-# Copyright (c) Microsoft Corporation. All rights reserved.
-# Licensed under the MIT License.
-set(SERVER_APP_NAME "onnxruntime_server")
-
-set(gRPC_BUILD_TESTS OFF CACHE INTERNAL "Don't build tests")
-set(gRPC_GFLAGS_PROVIDER "" CACHE INTERNAL "Don't use gflags")
-set(gRPC_BENCHMARK_PROVIDER "" CACHE INTERNAL "Don't use benchmark")
-set(gRPC_ZLIB_PROVIDER "package" CACHE INTERNAL "Use preinstalled zlib library")
-set(gRPC_PROTOBUF_PROVIDER "" CACHE INTERNAL "Don't use grpc protobuf, set it manually.")
-
-
-# protobuf targets have already been included as submodules - adapted from https://github.com/grpc/grpc/blob/master/cmake/protobuf.cmake
-set(_gRPC_PROTOBUF_LIBRARY_NAME "libprotobuf")
-set(_gRPC_PROTOBUF_LIBRARIES protobuf::${_gRPC_PROTOBUF_LIBRARY_NAME})
-
-set(_gRPC_PROTOBUF_PROTOC_LIBRARIES protobuf::libprotoc)
-# extract the include dir from target's properties
-
-set(_gRPC_PROTOBUF_WELLKNOWN_INCLUDE_DIR ${REPO_ROOT}/cmake/external/protobuf/src)
-set(_gRPC_PROTOBUF_PROTOC protobuf::protoc)
-set(_gRPC_PROTOBUF_PROTOC_EXECUTABLE $<TARGET_FILE:protobuf::protoc>)
-
-set(_gRPC_PROTOBUF_INCLUDE_DIR ${PROTOBUF_INCLUDE_DIRS})
-
-if(NOT WIN32)
-  string(REPLACE "-Werror" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") # Disable werror for included subdirectories - c-ares<1.15 breaks with -Wall
-  string(REPLACE "-Werror" "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}")
-  if(HAS_UNUSED_PARAMETER) # disable warning for unused parameters because (BoringSSL specifically) have unused parameters.
-    string(APPEND CMAKE_CXX_FLAGS " -Wno-unused-parameter")
-    string(APPEND CMAKE_C_FLAGS " -Wno-unused-parameter")
-  endif()
-endif()
-
-add_subdirectory(${PROJECT_SOURCE_DIR}/external/grpc EXCLUDE_FROM_ALL)
-if(NOT WIN32)
-  if(onnxruntime_DEV_MODE) # Reenable Werror for our code subdirectories.
-      if(NOT onnxruntime_USE_TVM)
-        string(APPEND CMAKE_CXX_FLAGS " -Werror")
-        string(APPEND CMAKE_C_FLAGS " -Werror")
-      endif()
-  endif()
-  if(HAS_UNUSED_PARAMETER) # reenable warning for unused parameters for our code.
-    string(APPEND CMAKE_CXX_FLAGS " -Wunused-parameter")
-    string(APPEND CMAKE_C_FLAGS " -Wunused-parameter")
-  endif()
-endif()
-
-set(_GRPC_CPP_PLUGIN_EXECUTABLE $<TARGET_FILE:grpc_cpp_plugin>)
-set(_GRPC_PY_PLUGIN_EXECUTABLE $<TARGET_FILE:grpc_python_plugin>)
-
-
-# Generate .h and .cc files from protobuf file
-add_library(server_proto ${ONNXRUNTIME_ROOT}/server/protobuf/predict.proto ${ONNXRUNTIME_ROOT}/server/protobuf/onnx-ml.proto)
-if(WIN32)
-  target_compile_options(server_proto PRIVATE "/wd4125" "/wd4456")
-endif()
-target_include_directories(server_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_INCLUDE_DIRECTORIES> "${CMAKE_CURRENT_BINARY_DIR}/.." ${CMAKE_CURRENT_BINARY_DIR}/onnx)
-target_compile_definitions(server_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_COMPILE_DEFINITIONS>)
-onnxruntime_protobuf_generate(APPEND_PATH IMPORT_DIRS ${REPO_ROOT}/cmake/external/protobuf/src ${ONNXRUNTIME_ROOT}/server/protobuf ${ONNXRUNTIME_ROOT}/core/protobuf TARGET server_proto)
-add_dependencies(server_proto ${onnxruntime_EXTERNAL_DEPENDENCIES})
-if(NOT WIN32)
-  if(HAS_UNUSED_PARAMETER)
-     set_source_files_properties(${CMAKE_CURRENT_BINARY_DIR}/predict.pb.cc PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-     set_source_files_properties(${CMAKE_CURRENT_BINARY_DIR}/onnx-ml.pb.cc PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-  endif()
-endif()
-
-# Setup dependencies
-include(get_boost.cmake)
-set(re2_src ${REPO_ROOT}/cmake/external/re2)
-set(SPDLOG_BUILD_EXAMPLES OFF)
-add_subdirectory(${REPO_ROOT}/cmake/external/spdlog)
-
-# Generate GRPC service source and headers.
-get_filename_component(grpc_proto "${ONNXRUNTIME_ROOT}/server/protobuf/prediction_service.proto" ABSOLUTE)
-get_filename_component(grpc_proto_path "${grpc_proto}" PATH)
-
-set(grpc_srcs "${CMAKE_CURRENT_BINARY_DIR}/prediction_service.grpc.pb.cc")
-set(grpc_hdrs "${CMAKE_CURRENT_BINARY_DIR}/prediction_service.grpc.pb.h")
-add_custom_command(
-      OUTPUT "${grpc_srcs}" "${grpc_hdrs}"
-      COMMAND $<TARGET_FILE:protobuf::protoc>
-      ARGS 
-        --cpp_out "${CMAKE_CURRENT_BINARY_DIR}"
-        --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
-        --plugin=protoc-gen-grpc="${_GRPC_CPP_PLUGIN_EXECUTABLE}"
-        -I ${grpc_proto_path}
-        "${grpc_proto}"
-      DEPENDS "${grpc_proto}" ${_GRPC_CPP_PLUGIN_EXECUTABLE}
-      COMMENT "Running ${_GRPC_CPP_PLUGIN_EXECUTABLE} on ${grpc_proto}"
-    )
-
-add_library(server_grpc_proto ${grpc_srcs})
-target_include_directories(server_grpc_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_INCLUDE_DIRECTORIES> "${CMAKE_CURRENT_BINARY_DIR}" ${CMAKE_CURRENT_BINARY_DIR}/onnx PRIVATE)
-if(APPLE)
-  set(grpc_reflection -Wl,-all_load grpc++_reflection -Wl,-noall_load)
-else()
-  set(grpc_reflection -Wl,--whole-archive grpc++_reflection -Wl,--no-whole-archive)
-endif()
-set(grpc_static_libs grpc++ grpcpp_channelz)
-target_link_libraries(server_grpc_proto ${grpc_static_libs})
-add_dependencies(server_grpc_proto server_proto)
-# Include generated *.pb.h files
-include_directories("${CMAKE_CURRENT_BINARY_DIR}")
-
-if(NOT WIN32)
-  if(HAS_UNUSED_PARAMETER)
-    set_source_files_properties(${grpc_srcs} PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-    set_source_files_properties(${onnxruntime_server_grpc_srcs} PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-  endif()
-endif()
-
-
-# Setup source code
-set(onnxruntime_server_lib_srcs
-  "${ONNXRUNTIME_ROOT}/server/http/json_handling.cc"
-  "${ONNXRUNTIME_ROOT}/server/http/predict_request_handler.cc"
-  "${ONNXRUNTIME_ROOT}/server/http/util.cc"
-  "${ONNXRUNTIME_ROOT}/server/environment.cc"
-  "${ONNXRUNTIME_ROOT}/server/executor.cc"
-  "${ONNXRUNTIME_ROOT}/server/converter.cc"
-  "${ONNXRUNTIME_ROOT}/server/util.cc"
-  "${ONNXRUNTIME_ROOT}/server/core/request_id.cc"
-  "${ONNXRUNTIME_ROOT}/server/grpc/prediction_service_impl.cc"
-  "${ONNXRUNTIME_ROOT}/server/grpc/grpc_app.cc"
-  "${ONNXRUNTIME_ROOT}/server/serializing/tensorprotoutils.cc"
-  )
-if(NOT WIN32)
-  if(HAS_UNUSED_PARAMETER)
-    set_source_files_properties(${onnxruntime_server_lib_srcs} PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-  endif()
-endif()
-
-file(GLOB_RECURSE onnxruntime_server_http_core_lib_srcs
-  "${ONNXRUNTIME_ROOT}/server/http/core/*.cc"
-  )
-
-file(GLOB_RECURSE onnxruntime_server_srcs
-  "${ONNXRUNTIME_ROOT}/server/main.cc"
-)
-
-# HTTP core library
-add_library(onnxruntime_server_http_core_lib STATIC
-  ${onnxruntime_server_http_core_lib_srcs})
-target_include_directories(onnxruntime_server_http_core_lib
-  PUBLIC
-  ${ONNXRUNTIME_ROOT}
-  ${ONNXRUNTIME_ROOT}/server/http/core
-  ${ONNXRUNTIME_ROOT}/server/core
-  ${Boost_INCLUDE_DIR}
-  ${re2_src}
-)
-add_dependencies(onnxruntime_server_http_core_lib Boost)
-target_link_libraries(onnxruntime_server_http_core_lib PRIVATE
-  ${Boost_LIBRARIES}
-)
-
-# Server library
-add_library(onnxruntime_server_lib ${onnxruntime_server_lib_srcs})
-onnxruntime_add_include_to_target(onnxruntime_server_lib onnx_proto server_proto)
-target_include_directories(onnxruntime_server_lib PRIVATE
-  ${ONNXRUNTIME_INCLUDE_DIR}
-  ${ONNXRUNTIME_ROOT}
-  ${ONNXRUNTIME_ROOT}/server
-  ${ONNXRUNTIME_ROOT}/server/http
-  ${ONNXRUNTIME_ROOT}/server/logging
-  ${ONNXRUNTIME_ROOT}/server/core
-  PUBLIC
-  ${ONNXRUNTIME_ROOT}/server
-  ${Boost_INCLUDE_DIR}
-  ${re2_src}
-)
-
-
-target_link_libraries(onnxruntime_server_lib PRIVATE
-  server_proto
-  server_grpc_proto
-  ${Boost_LIBRARIES}
-  onnxruntime_server_http_core_lib
-  PUBLIC
-  protobuf::libprotobuf
-  ${onnxruntime_EXTERNAL_LIBRARIES}
-  spdlog::spdlog
-  onnxruntime
-)
-
-if (onnxruntime_USE_SYSLOG)
-  target_compile_definitions(onnxruntime_server_lib PUBLIC USE_SYSLOG="1")
-endif()
-
-# For IDE only
-source_group(TREE ${REPO_ROOT} FILES ${onnxruntime_server_srcs} ${onnxruntime_server_lib_srcs} ${onnxruntime_server_lib})
-
-# Server Application
-add_executable(${SERVER_APP_NAME} ${onnxruntime_server_srcs})
-add_dependencies(${SERVER_APP_NAME} onnx server_proto onnx_proto ${onnxruntime_EXTERNAL_DEPENDENCIES})
-
-if(NOT WIN32)
-  if(HAS_UNUSED_PARAMETER)
-    set_source_files_properties("${ONNXRUNTIME_ROOT}/server/main.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-  endif()
-endif()
-
-set(onnxruntime_SERVER_VERSION "local-build" CACHE STRING "Sever version")
-target_compile_definitions(${SERVER_APP_NAME} PUBLIC SRV_VERSION="${onnxruntime_SERVER_VERSION}")
-message(STATUS "ONNX Runtime Server version set to: ${onnxruntime_SERVER_VERSION}")
-
-set(onnxruntime_LATEST_COMMIT_ID "default" CACHE STRING "The latest commit id")
-target_compile_definitions(${SERVER_APP_NAME} PUBLIC LATEST_COMMIT_ID="${onnxruntime_LATEST_COMMIT_ID}")
-message(STATUS "ONNX Runtime Server latest commit id is: ${onnxruntime_LATEST_COMMIT_ID}")
-
-onnxruntime_add_include_to_target(${SERVER_APP_NAME} onnxruntime_session onnxruntime_server_lib onnx onnx_proto server_proto)
-
-target_include_directories(${SERVER_APP_NAME} PRIVATE
-    ${ONNXRUNTIME_INCLUDE_DIR}
-    ${ONNXRUNTIME_ROOT}/server/http
-)
-
-
-target_link_libraries(${SERVER_APP_NAME} PRIVATE
-    onnxruntime_server_http_core_lib
-    onnxruntime_server_lib
-    ${grpc_reflection} #Note that this will break the tests if we try to link it to the lib so just link to the executable.
-)
-
diff --git a/cmake/onnxruntime_unittests.cmake b/cmake/onnxruntime_unittests.cmake
index 44dbd0017c710..d3564feffb719 100644
--- a/cmake/onnxruntime_unittests.cmake
+++ b/cmake/onnxruntime_unittests.cmake
@@ -698,87 +698,6 @@ if (onnxruntime_BUILD_SHARED_LIB)
   )
 endif()
 
-if (onnxruntime_BUILD_SERVER)
-  file(GLOB onnxruntime_test_server_src
-    "${TEST_SRC_DIR}/server/unit_tests/*.cc"
-    "${TEST_SRC_DIR}/server/unit_tests/*.h"
-  )
-
-  file(GLOB onnxruntime_integration_test_server_src
-    "${TEST_SRC_DIR}/server/integration_tests/*.py"
-  )
-  if(NOT WIN32)
-    if(HAS_UNUSED_PARAMETER)
-      set_source_files_properties("${TEST_SRC_DIR}/server/unit_tests/json_handling_tests.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-      set_source_files_properties("${TEST_SRC_DIR}/server/unit_tests/converter_tests.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-      set_source_files_properties("${TEST_SRC_DIR}/server/unit_tests/util_tests.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-      set_source_files_properties("${TEST_SRC_DIR}/server/unit_tests/prediction_service_impl_test.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-      set_source_files_properties("${TEST_SRC_DIR}/server/unit_tests/executor_test.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
-    endif()
-  endif()
-  
-  add_library(onnxruntime_test_utils_for_server ${onnxruntime_test_server_src})
-  onnxruntime_add_include_to_target(onnxruntime_test_utils_for_server onnxruntime_test_utils_for_framework gtest gmock onnx onnx_proto server_proto server_grpc_proto)
-  add_dependencies(onnxruntime_test_utils_for_server onnxruntime_server_lib onnxruntime_server_http_core_lib Boost ${onnxruntime_EXTERNAL_DEPENDENCIES})
-  target_include_directories(onnxruntime_test_utils_for_server PUBLIC ${Boost_INCLUDE_DIR} ${REPO_ROOT}/cmake/external/re2 ${CMAKE_CURRENT_BINARY_DIR}/onnx ${ONNXRUNTIME_ROOT}/server ${ONNXRUNTIME_ROOT}/server/http ${ONNXRUNTIME_ROOT}/server/http/core  ${ONNXRUNTIME_ROOT}/server/grpc ${ONNXRUNTIME_ROOT}/server ${ONNXRUNTIME_ROOT}/server/core PRIVATE ${ONNXRUNTIME_ROOT})
- if (onnxruntime_USE_OPENVINO)
-   message(${OPENVINO_INCLUDE_DIR})
-   target_include_directories(onnxruntime_test_utils_for_server PUBLIC ${OPENVINO_INCLUDE_DIR} ${OPENVINO_TBB_INCLUDE_DIR})
- endif()
-  if(UNIX)
-    target_compile_options(onnxruntime_test_utils_for_server PRIVATE "$<$<COMPILE_LANGUAGE:CUDA>:SHELL:-Xcompiler -Wno-error=sign-compare>"
-            "$<$<NOT:$<COMPILE_LANGUAGE:CUDA>>:-Wno-error=sign-compare>")
-  endif()
-  target_link_libraries(onnxruntime_test_utils_for_server ${Boost_LIBRARIES} spdlog::spdlog server_grpc_proto)
-
-
-  AddTest(
-    TARGET onnxruntime_server_tests
-    SOURCES ${onnxruntime_test_server_src}
-    LIBS ${onnxruntime_test_server_libs} server_proto server_grpc_proto onnxruntime_server_lib ${onnxruntime_test_providers_libs}
-    DEPENDS ${onnxruntime_EXTERNAL_DEPENDENCIES}
-  )
-
-  onnxruntime_protobuf_generate(
-          APPEND_PATH IMPORT_DIRS ${REPO_ROOT}/cmake/external/protobuf/src ${ONNXRUNTIME_ROOT}/server/protobuf ${ONNXRUNTIME_ROOT}/core/protobuf
-          PROTOS ${ONNXRUNTIME_ROOT}/server/protobuf/predict.proto ${ONNXRUNTIME_ROOT}/server/protobuf/onnx-ml.proto
-          LANGUAGE python
-          TARGET onnxruntime_server_tests
-          OUT_VAR server_test_py)
-          
-  set(grpc_py "${CMAKE_CURRENT_BINARY_DIR}/prediction_service_pb2_grpc.py")
-
-  add_custom_command(
-    TARGET onnxruntime_server_tests
-    COMMAND $<TARGET_FILE:protobuf::protoc>
-    ARGS 
-      --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
-      --plugin=protoc-gen-grpc="${_GRPC_PY_PLUGIN_EXECUTABLE}"
-      -I ${grpc_proto_path}
-      "${grpc_proto}"
-    DEPENDS "${grpc_proto}"
-    COMMENT "Running ${_GRPC_PY_PLUGIN_EXECUTABLE} on ${grpc_proto}"
-    )
-
-  add_custom_command(
-    TARGET onnxruntime_server_tests POST_BUILD
-    COMMAND ${CMAKE_COMMAND} -E make_directory ${CMAKE_CURRENT_BINARY_DIR}/server_test
-    COMMAND ${CMAKE_COMMAND} -E copy
-      ${onnxruntime_integration_test_server_src}
-      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
-      COMMAND ${CMAKE_COMMAND} -E copy
-      ${CMAKE_CURRENT_BINARY_DIR}/onnx_ml_pb2.py
-      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
-    COMMAND ${CMAKE_COMMAND} -E copy
-      ${CMAKE_CURRENT_BINARY_DIR}/predict_pb2.py
-      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
-    COMMAND ${CMAKE_COMMAND} -E copy
-      ${grpc_py}
-      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
-  )
-
-endif()
-
 #some ETW tools
 if(WIN32 AND onnxruntime_ENABLE_INSTRUMENT)
     add_executable(generate_perf_report_from_etl ${ONNXRUNTIME_ROOT}/tool/etw/main.cc ${ONNXRUNTIME_ROOT}/tool/etw/eparser.h ${ONNXRUNTIME_ROOT}/tool/etw/eparser.cc ${ONNXRUNTIME_ROOT}/tool/etw/TraceSession.h ${ONNXRUNTIME_ROOT}/tool/etw/TraceSession.cc)
diff --git a/onnxruntime/test/server/unit_tests/converter_tests.cc b/onnxruntime/test/server/unit_tests/converter_tests.cc
deleted file mode 100644
index 6387391bbf98f..0000000000000
--- a/onnxruntime/test/server/unit_tests/converter_tests.cc
+++ /dev/null
@@ -1,1113 +0,0 @@
-// Copyright (c) Microsoft Corporation. All rights reserved.
-// Licensed under the MIT License.
-
-#include "gtest/gtest.h"
-#include "gmock/gmock.h"
-
-#include "core/framework/tensor.h"
-#include "core/graph/basic_types.h"
-#include "core/framework/allocatormgr.h"
-#include "test/framework/test_utils.h"
-#include "test/test_environment.h"
-#include "server/converter.h"
-#include "server/serializing/tensorprotoutils.h"
-#include <spdlog/spdlog.h>
-#include <spdlog/sinks/sink.h>
-#include <spdlog/sinks/null_sink.h>
-#include "onnx-ml.pb.h"
-
-namespace onnxruntime {
-namespace server {
-namespace test {
-
-void CreateMLValueBool(AllocatorPtr alloc, const std::vector<int64_t>& dims, const bool* value, Ort::Value& p_mlvalue);
-
-template <typename T>
-void CreateMLValue(AllocatorPtr alloc, const std::vector<int64_t>& dims, const std::vector<T>& value,
-                   Ort::Value& p_mlvalue) {
-  OrtValue* ml_value = new OrtValue{};
-  onnxruntime::test::CreateMLValue<T>(alloc, dims, value, ml_value);
-  p_mlvalue = Ort::Value{ml_value};
-}
-
-IExecutionProvider* TestCPUExecutionProvider() {
-  static CPUExecutionProviderInfo info;
-  static CPUExecutionProvider cpu_provider(info);
-  return &cpu_provider;
-}
-
-TEST(MLDataTypeToTensorProtoDataTypeTests, MLDataTypeToTensorProtoDataTypeTests) {
-  onnx::TensorProto_DataType result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_FLOAT);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT16);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_FLOAT16);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_BFLOAT16);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_BFLOAT16);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_DOUBLE);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_DOUBLE);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT8);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_UINT8);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_INT8);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_INT8);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT16);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_UINT16);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_INT16);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_INT16);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT32);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_UINT32);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_INT32);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_INT32);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT64);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_UINT64);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_INT64);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_STRING);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_BOOL);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_BOOL);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(ONNX_TENSOR_ELEMENT_DATA_TYPE_UNDEFINED);
-  EXPECT_EQ(result, onnx::TensorProto_DataType_UNDEFINED);
-
-  result = onnxruntime::server::MLDataTypeToTensorProtoDataType(static_cast<ONNXTensorElementDataType>(17));
-  EXPECT_EQ(result, onnx::TensorProto_DataType_UNDEFINED);
-}
-
-TEST(MLValueToTensorProtoTests, FloatToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<float> values_mul_x = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<float>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_FLOAT);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(float);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const float*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_FLOAT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, FloatToFloatData) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<float> values_mul_x = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<float>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_FLOAT);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  EXPECT_EQ(tp.float_data().size(), 6);
-  auto data = tp.float_data().data();
-  for (int j = 0; j < tp.float_data().size(); ++j) {
-    EXPECT_FLOAT_EQ(data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int32ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int32_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int32_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT32);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(int32_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const int32_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int32ToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int32_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int32_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT32);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  EXPECT_EQ(tp.int32_data().size(), 6);
-  auto data = tp.int32_data().data();
-  for (int j = 0; j < tp.int32_data().size(); ++j) {
-    EXPECT_EQ(data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt8ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint8_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint8_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT8);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint8_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const uint8_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt8ToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint8_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint8_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT8);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 6);
-  for (int x = 0; x < 6; ++x) {
-    EXPECT_EQ(tp.int32_data()[x], values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt8ProtoRoundTrip) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint8_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-
-  onnx::TensorProto tp;
-  for (auto const& val : values_mul_x) {
-    tp.add_int32_data(val);
-  }
-  for (auto const& dim : dims_mul_x) {
-    tp.add_dims(dim);
-  }
-  tp.set_data_type(onnx::TensorProto_DataType_UINT8);
-  Ort::Value ml_value{nullptr};
-  char buf[1000];
-  Ort::AllocatorWithDefaultOptions allocator;
-  auto info = allocator.GetInfo();
-  MemBuffer buffer((void*)&buf, tp.ByteSizeLong(), *info);
-  onnxruntime::server::TensorProtoToMLValue(tp, buffer, ml_value);
-
-  onnx::TensorProto tp_out;
-
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp_out);
-
-  // Verify data type
-  EXPECT_TRUE(tp_out.has_data_type());
-  EXPECT_EQ(tp_out.data_type(), onnx::TensorProto_DataType_UINT8);
-
-  // Verify data location
-  EXPECT_FALSE(tp_out.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp_out.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp_out.has_raw_data());
-
-  EXPECT_EQ(tp_out.int32_data_size(), tp.int32_data_size());
-  auto in_data = tp.int32_data();
-  auto out_data = tp_out.int32_data();
-  for (auto x = 0; x < tp_out.int32_data_size(); ++x) {
-    EXPECT_EQ(values_mul_x[x], in_data[x]);
-    EXPECT_EQ(static_cast<uint8_t>(out_data[x]), in_data[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int8ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int8_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int8_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT8);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint8_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const int8_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int8ToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int8_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int8_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT8);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 6);
-  auto data = tp.int32_data();
-  for (int x = 0; x < 6; ++x) {
-    EXPECT_EQ(tp.int32_data()[x], values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt16ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 3};
-  std::vector<uint16_t> values_mul_x = {1, 2, 3, 4, 5, 6, 7, 8, 9};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint16_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT16);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint16_t);
-  EXPECT_EQ(count, 9);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const uint16_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt16ToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 3};
-  std::vector<uint16_t> values_mul_x = {1, 2, 3, 4, 5, 6, 7, 8, 9};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint16_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT16);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 9);
-  auto data = tp.int32_data();
-  for (int x = 0; x < 9; ++x) {
-    EXPECT_EQ(tp.int32_data()[x], values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int16ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int16_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int16_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT16);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint16_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const int16_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int16ToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int16_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int16_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT16);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 6);
-  auto data = tp.int32_data();
-  for (int x = 0; x < 6; ++x) {
-    EXPECT_EQ(tp.int32_data()[x], values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, BoolToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  bool values_mul_x[] = {true, false, false, true, true, false};
-  Ort::Value ml_value{nullptr};
-  CreateMLValueBool(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_BOOL);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(bool);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const bool*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, BoolToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  bool values_mul_x[] = {true, false, false, true, true, false};
-  Ort::Value ml_value{nullptr};
-  CreateMLValueBool(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_BOOL);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 6);
-  auto data = tp.int32_data();
-  for (int x = 0; x < 6; ++x) {
-    EXPECT_EQ(tp.int32_data()[x], values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Float16ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<onnxruntime::MLFloat16> values_mul_x{
-      onnxruntime::MLFloat16(1),
-      onnxruntime::MLFloat16(2),
-      onnxruntime::MLFloat16(3),
-      onnxruntime::MLFloat16(4),
-      onnxruntime::MLFloat16(5),
-      onnxruntime::MLFloat16(6)};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<onnxruntime::MLFloat16>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_FLOAT16);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(onnxruntime::MLFloat16);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const onnxruntime::MLFloat16*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, FloatToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<onnxruntime::MLFloat16> values_mul_x{
-      onnxruntime::MLFloat16(1),
-      onnxruntime::MLFloat16(2),
-      onnxruntime::MLFloat16(3),
-      onnxruntime::MLFloat16(4),
-      onnxruntime::MLFloat16(5),
-      onnxruntime::MLFloat16(6)};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<onnxruntime::MLFloat16>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_FLOAT16);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 6);
-  auto data = tp.int32_data().data();
-  for (int x = 0; x < 6; ++x) {
-    const u_int16_t data16 = data[x];
-    const auto data_float_16 = static_cast<onnxruntime::MLFloat16>(data16);
-    EXPECT_EQ(data_float_16, values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, BFloat16ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<onnxruntime::BFloat16> values_mul_x{
-      onnxruntime::BFloat16(1.0f),
-      onnxruntime::BFloat16(2.0f),
-      onnxruntime::BFloat16(3.0f),
-      onnxruntime::BFloat16(4.0f),
-      onnxruntime::BFloat16(5.0f),
-      onnxruntime::BFloat16(6.0f)};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<onnxruntime::BFloat16>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_BFLOAT16);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint16_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const uint16_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j].val);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, BFloatToInt32Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<onnxruntime::BFloat16> values_mul_x{
-      onnxruntime::BFloat16(1.0f),
-      onnxruntime::BFloat16(2.0f),
-      onnxruntime::BFloat16(3.0f),
-      onnxruntime::BFloat16(4.0f),
-      onnxruntime::BFloat16(5.0f),
-      onnxruntime::BFloat16(6.0f)};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<onnxruntime::BFloat16>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_BFLOAT16);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.int32_data().size();
-  EXPECT_EQ(count, 6);
-  auto data = tp.int32_data().data();
-  for (int x = 0; x < 6; ++x) {
-    const u_int16_t data16 = data[x];
-    const auto data_float_16 = static_cast<onnxruntime::BFloat16>(data16);
-    EXPECT_EQ(data_float_16, values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, StringToStringData) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<std::string> values_mul_x{"A", "BC", "DEF", "123", "45", "6"};
-  OrtValue* p_mlValue = new OrtValue{};
-  onnxruntime::test::AllocateMLValue<std::string>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, p_mlValue);
-
-  Tensor* mutable_tensor = p_mlValue->GetMutable<Tensor>();
-  std::string* mutable_data = mutable_tensor->MutableData<std::string>();
-  for (size_t i = 0; i < values_mul_x.size(); ++i) {
-    mutable_data[i] = values_mul_x[i];
-  }
-  Ort::Value value{p_mlValue};
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_STRING);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.string_data().size();
-  EXPECT_EQ(count, 6);
-  const auto* data = tp.string_data().data();
-  for (int x = 0; x < 6; ++x) {
-    EXPECT_STREQ(data[x]->c_str(), values_mul_x[x].c_str());
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int64ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int64_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int64_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT64);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(int64_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const int64_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, Int64ToInt64Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<int64_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<int64_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_INT64);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  EXPECT_EQ(tp.int64_data().size(), 6);
-  auto data = tp.int64_data().data();
-  for (int j = 0; j < tp.int64_data().size(); ++j) {
-    EXPECT_EQ(data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt32ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint32_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint32_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT32);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint32_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  auto* tensor_data = (uint32_t*)raw;
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt32ToUint64Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint32_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint32_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT32);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  auto count = tp.uint64_data().size();
-  EXPECT_EQ(count, 6);
-
-  auto data = tp.uint64_data().data();
-  for (size_t x = 0; x < count; ++x) {
-    EXPECT_EQ(data[x], values_mul_x[x]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt64ToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint64_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint64_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT64);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint64_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const uint64_t*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, UInt64ToInt64Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<uint64_t> values_mul_x = {1, 2, 3, 4, 5, 6};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<uint64_t>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_UINT64);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  EXPECT_EQ(tp.uint64_data().size(), 6);
-  auto data = tp.uint64_data().data();
-  for (int j = 0; j < tp.uint64_data().size(); ++j) {
-    EXPECT_EQ(data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, DoubleToRaw) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<double> values_mul_x = {1.0, 2.0, 3.0, 4.0, 5.0, 6.0};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<double>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ true, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_DOUBLE);
-
-  // Verify data location
-  EXPECT_TRUE(tp.has_data_location());
-  EXPECT_EQ(tp.data_location(), onnx::TensorProto_DataLocation_DEFAULT);
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_TRUE(tp.has_raw_data());
-  auto count = tp.raw_data().size() / sizeof(uint64_t);
-  EXPECT_EQ(count, 6);
-
-  auto raw = tp.raw_data().data();
-  const auto* tensor_data = reinterpret_cast<const double*>(raw);
-  for (size_t j = 0; j < count; ++j) {
-    EXPECT_DOUBLE_EQ(tensor_data[j], values_mul_x[j]);
-  }
-}
-
-TEST(MLValueToTensorProtoTests, DoubleToInt64Data) {
-  std::vector<int64_t> dims_mul_x = {3, 2};
-  std::vector<double> values_mul_x = {1.0, 2.0, 3.0, 4.0, 5.0, 6.0};
-  Ort::Value ml_value{nullptr};
-  onnxruntime::server::test::CreateMLValue<double>(TestCPUExecutionProvider()->GetAllocator(0, OrtMemTypeDefault), dims_mul_x, values_mul_x, ml_value);
-
-  onnx::TensorProto tp;
-  onnxruntime::server::MLValueToTensorProto(ml_value, /* using_raw_data */ false, spdlog::default_logger(), tp);
-
-  // Verify data type
-  EXPECT_TRUE(tp.has_data_type());
-  EXPECT_EQ(tp.data_type(), onnx::TensorProto_DataType_DOUBLE);
-
-  // Verify data location
-  EXPECT_FALSE(tp.has_data_location());
-
-  // Verify dimensions
-  const auto& dims = tp.dims();
-  std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
-  for (int i = 0; i < dims.size(); ++i) {
-    EXPECT_EQ(dims[i], dims_mul_x[i]);
-  }
-
-  // Verify data
-  EXPECT_FALSE(tp.has_raw_data());
-  EXPECT_EQ(tp.double_data().size(), 6);
-  auto data = tp.double_data().data();
-  for (int j = 0; j < tp.double_data().size(); ++j) {
-    EXPECT_DOUBLE_EQ(data[j], values_mul_x[j]);
-  }
-}
-
-void CreateMLValueBool(AllocatorPtr alloc, const std::vector<int64_t>& dims, const bool* value, Ort::Value& p_value) {
-  TensorShape shape(dims);
-  OrtValue* p_mlvalue = new OrtValue{};
-  auto element_type = DataTypeImpl::GetType<bool>();
-  std::unique_ptr<Tensor> p_tensor = onnxruntime::make_unique<Tensor>(element_type,
-                                                              shape,
-                                                              alloc);
-  memcpy(p_tensor->MutableData<bool>(), &value[0], element_type->Size() * shape.Size());
-  p_mlvalue->Init(p_tensor.release(),
-                  DataTypeImpl::GetType<Tensor>(),
-                  DataTypeImpl::GetType<Tensor>()->GetDeleteFunc());
-  p_value = Ort::Value{p_mlvalue};
-}
-
-}  // namespace test
-}  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
diff --git a/server/CMakeLists.txt b/server/CMakeLists.txt
new file mode 100755
index 0000000000000..efefb84aec691
--- /dev/null
+++ b/server/CMakeLists.txt
@@ -0,0 +1,258 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+# Minimum CMake required
+cmake_minimum_required(VERSION 3.5)
+# Project
+project(onnxruntime C CXX)
+set(CMAKE_CXX_STANDARD 14)
+
+find_package(Threads)
+list(APPEND CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake)
+include(protobuf_function.cmake)
+find_package(gRPC REQUIRED)
+
+set(SERVER_APP_NAME "onnxruntime_server")
+set(onnxruntime_USE_FULL_PROTOBUF ON)
+
+set(ONNXRUNTIME_SERVER_ROOT ${PROJECT_SOURCE_DIR})
+
+# Generate .h and .cc files from protobuf file
+add_library(server_proto ${ONNXRUNTIME_SERVER_ROOT}/protobuf/predict.proto ${ONNXRUNTIME_SERVER_ROOT}/protobuf/onnx-ml.proto)
+if(WIN32)
+  target_compile_options(server_proto PRIVATE "/wd4125" "/wd4456")
+endif()
+target_include_directories(server_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_INCLUDE_DIRECTORIES> "${CMAKE_CURRENT_BINARY_DIR}/.." ${CMAKE_CURRENT_BINARY_DIR}/onnx)
+target_compile_definitions(server_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_COMPILE_DEFINITIONS>)
+onnxruntime_protobuf_generate(APPEND_PATH ${ONNXRUNTIME_SERVER_ROOT}/protobuf ${ONNXRUNTIME_ROOT}/core/protobuf TARGET server_proto)
+if(NOT WIN32)
+  if(HAS_UNUSED_PARAMETER)
+     set_source_files_properties(${CMAKE_CURRENT_BINARY_DIR}/predict.pb.cc PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+     set_source_files_properties(${CMAKE_CURRENT_BINARY_DIR}/onnx-ml.pb.cc PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+  endif()
+endif()
+
+# Setup dependencies
+include(get_boost.cmake)
+set(SPDLOG_BUILD_EXAMPLES OFF)
+add_subdirectory(external/spdlog)
+
+# Generate GRPC service source and headers.
+get_filename_component(grpc_proto "${ONNXRUNTIME_SERVER_ROOT}/protobuf/prediction_service.proto" ABSOLUTE)
+get_filename_component(grpc_proto_path "${grpc_proto}" PATH)
+
+set(grpc_srcs "${CMAKE_CURRENT_BINARY_DIR}/prediction_service.grpc.pb.cc")
+set(grpc_hdrs "${CMAKE_CURRENT_BINARY_DIR}/prediction_service.grpc.pb.h")
+
+if(NOT _gRPC_PYTHON_PLUGIN_EXECUTABLE)
+  find_program(_gRPC_PYTHON_PLUGIN_EXECUTABLE grpc_python_plugin DOC "The gRPC Python plugin for protoc")
+endif()
+
+add_custom_command(
+      OUTPUT "${grpc_srcs}" "${grpc_hdrs}"
+      COMMAND protoc
+      ARGS 
+        --cpp_out "${CMAKE_CURRENT_BINARY_DIR}"
+        --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
+        --plugin=protoc-gen-grpc="${_gRPC_CPP_PLUGIN_EXECUTABLE}"
+        -I ${grpc_proto_path}
+        "${grpc_proto}"
+      DEPENDS "${grpc_proto}"
+      COMMENT "Running ${_gRPC_CPP_PLUGIN_EXECUTABLE} on ${grpc_proto}"
+    )
+
+add_library(server_grpc_proto ${grpc_srcs})
+target_include_directories(server_grpc_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_INCLUDE_DIRECTORIES> "${CMAKE_CURRENT_BINARY_DIR}" ${CMAKE_CURRENT_BINARY_DIR}/onnx PRIVATE)
+set(grpc_reflection -Wl,--whole-archive grpc++_reflection -Wl,--no-whole-archive)
+set(grpc_static_libs gRPC::grpc++ grpcpp_channelz)
+
+add_dependencies(server_grpc_proto server_proto Boost)
+# Include generated *.pb.h files
+include_directories("${CMAKE_CURRENT_BINARY_DIR}")
+
+if(NOT WIN32)
+  if(HAS_UNUSED_PARAMETER)
+    set_source_files_properties(${grpc_srcs} PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+    set_source_files_properties(${onnxruntime_server_grpc_srcs} PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+  endif()
+endif()
+
+
+# Setup source code
+set(onnxruntime_server_lib_srcs
+  "${ONNXRUNTIME_SERVER_ROOT}/http/json_handling.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/http/predict_request_handler.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/http/util.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/environment.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/executor.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/converter.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/util.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/core/request_id.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/grpc/prediction_service_impl.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/grpc/grpc_app.cc"
+  "${ONNXRUNTIME_SERVER_ROOT}/serializing/tensorprotoutils.cc"
+  )
+if(NOT WIN32)
+  if(HAS_UNUSED_PARAMETER)
+    set_source_files_properties(${onnxruntime_server_lib_srcs} PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+  endif()
+endif()
+
+file(GLOB_RECURSE onnxruntime_server_http_core_lib_srcs
+  "${ONNXRUNTIME_SERVER_ROOT}/http/core/*.cc"
+  )
+
+file(GLOB_RECURSE onnxruntime_server_srcs
+  "${ONNXRUNTIME_SERVER_ROOT}/main.cc"
+)
+
+
+
+
+# HTTP core library
+add_library(onnxruntime_server_http_core_lib STATIC
+  ${onnxruntime_server_http_core_lib_srcs})
+target_include_directories(onnxruntime_server_http_core_lib
+  PUBLIC
+  ${ONNXRUNTIME_ROOT}
+  ${ONNXRUNTIME_SERVER_ROOT}/http/core
+  ${ONNXRUNTIME_SERVER_ROOT}/core
+  ${Boost_INCLUDE_DIR}
+  
+)
+add_dependencies(onnxruntime_server_http_core_lib Boost)
+
+# Server library
+add_library(onnxruntime_server_lib ${onnxruntime_server_lib_srcs})
+target_include_directories(onnxruntime_server_lib PRIVATE
+  ${ONNXRUNTIME_INCLUDE_DIR}
+  ${ONNXRUNTIME_ROOT}
+  ${ONNXRUNTIME_SERVER_ROOT}
+  ${ONNXRUNTIME_SERVER_ROOT}/http
+  ${ONNXRUNTIME_SERVER_ROOT}/logging
+  ${ONNXRUNTIME_SERVER_ROOT}/core  
+  PUBLIC
+  ${ONNXRUNTIME_SERVER_ROOT}/external/spdlog/include
+  ${ONNXRUNTIME_SERVER_ROOT}/http/core
+  ${ONNXRUNTIME_SERVER_ROOT}
+  ${Boost_INCLUDE_DIR}
+)
+
+
+if (onnxruntime_USE_SYSLOG)
+  target_compile_definitions(onnxruntime_server_lib PUBLIC USE_SYSLOG="1")
+endif()
+add_dependencies(onnxruntime_server_lib server_proto Boost)
+
+# Server Application
+add_executable(${SERVER_APP_NAME} ${onnxruntime_server_srcs})
+add_dependencies(${SERVER_APP_NAME} server_proto Boost)
+
+if(NOT WIN32)
+  if(HAS_UNUSED_PARAMETER)
+    set_source_files_properties("${ONNXRUNTIME_SERVER_ROOT}/main.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+  endif()
+endif()
+
+set(onnxruntime_SERVER_VERSION "local-build" CACHE STRING "Sever version")
+target_compile_definitions(${SERVER_APP_NAME} PUBLIC SRV_VERSION="${onnxruntime_SERVER_VERSION}")
+message(STATUS "ONNX Runtime Server version set to: ${onnxruntime_SERVER_VERSION}")
+
+set(onnxruntime_LATEST_COMMIT_ID "default" CACHE STRING "The latest commit id")
+target_compile_definitions(${SERVER_APP_NAME} PUBLIC LATEST_COMMIT_ID="${onnxruntime_LATEST_COMMIT_ID}")
+message(STATUS "ONNX Runtime Server latest commit id is: ${onnxruntime_LATEST_COMMIT_ID}")
+
+target_include_directories(${SERVER_APP_NAME} PRIVATE
+    ${ONNXRUNTIME_INCLUDE_DIR}
+    ${ONNXRUNTIME_SERVER_ROOT}/http
+)
+
+
+target_link_libraries(${SERVER_APP_NAME} PRIVATE
+    onnxruntime_server_http_core_lib
+    onnxruntime_server_lib
+    ${grpc_reflection} #Note that this will break the tests if we try to link it to the lib so just link to the executable.   
+   server_grpc_proto server_proto
+  ${Boost_LIBRARIES}
+  onnxruntime_server_http_core_lib  
+  ${grpc_static_libs}  protobuf::libprotobuf
+  spdlog::spdlog
+  onnxruntime gtest re2  ${CMAKE_DL_LIBS} Threads::Threads
+)
+
+
+
+
+file(GLOB onnxruntime_test_server_src
+    "test/unit_tests/*.cc"
+    "test/unit_tests/*.h"
+  )
+
+  file(GLOB onnxruntime_integration_test_server_src
+    "test/integration_tests/*.py"
+  )
+  if(NOT WIN32)
+    if(HAS_UNUSED_PARAMETER)
+      set_source_files_properties("test/unit_tests/json_handling_tests.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+      set_source_files_properties("test/unit_tests/converter_tests.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+      set_source_files_properties("test/unit_tests/util_tests.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+      set_source_files_properties("test/unit_tests/prediction_service_impl_test.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+      set_source_files_properties("test/unit_tests/executor_test.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+    endif()
+  endif()
+  
+  add_executable(onnxruntime_server_tests ${onnxruntime_test_server_src})
+  add_dependencies(onnxruntime_server_tests server_proto Boost)
+  target_link_libraries(onnxruntime_server_tests PRIVATE onnxruntime_server_http_core_lib
+    onnxruntime_server_lib
+    ${grpc_reflection} #Note that this will break the tests if we try to link it to the lib so just link to the executable.   
+   server_grpc_proto server_proto
+  ${Boost_LIBRARIES}
+  onnxruntime_server_http_core_lib  
+  ${grpc_static_libs} protobuf::libprotobuf
+  spdlog::spdlog
+  onnxruntime gtest re2 gtest ${CMAKE_DL_LIBS} Threads::Threads)
+  target_include_directories(onnxruntime_server_tests PRIVATE ${ONNXRUNTIME_SERVER_ROOT}/external/spdlog/include
+  ${ONNXRUNTIME_SERVER_ROOT}/http/core
+  ${ONNXRUNTIME_SERVER_ROOT} ${ONNXRUNTIME_SERVER_ROOT}/core)
+  onnxruntime_protobuf_generate(
+          APPEND_PATH IMPORT_DIRS ${ONNXRUNTIME_SERVER_ROOT}/protobuf 
+          PROTOS ${ONNXRUNTIME_SERVER_ROOT}/protobuf/predict.proto protobuf/onnx-ml.proto
+          LANGUAGE python
+          TARGET onnxruntime_server_tests
+          OUT_VAR server_test_py)
+          
+  set(grpc_py "${CMAKE_CURRENT_BINARY_DIR}/prediction_service_pb2_grpc.py")
+
+  add_custom_command(
+    TARGET onnxruntime_server_tests
+    COMMAND protoc
+    ARGS 
+      --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
+      --plugin=protoc-gen-grpc=${_gRPC_PYTHON_PLUGIN_EXECUTABLE}
+      -I ${grpc_proto_path}
+      "${grpc_proto}"
+    DEPENDS "${grpc_proto}"
+    COMMENT "Running ${_gRPC_PYTHON_PLUGIN_EXECUTABLE} on ${grpc_proto}"
+    )
+
+  add_custom_command(
+    TARGET onnxruntime_server_tests POST_BUILD
+    COMMAND ${CMAKE_COMMAND} -E make_directory ${CMAKE_CURRENT_BINARY_DIR}/server_test
+    COMMAND ${CMAKE_COMMAND} -E copy
+      ${onnxruntime_integration_test_server_src}
+      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
+      COMMAND ${CMAKE_COMMAND} -E copy
+      ${CMAKE_CURRENT_BINARY_DIR}/onnx_ml_pb2.py
+      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
+    COMMAND ${CMAKE_COMMAND} -E copy
+      ${CMAKE_CURRENT_BINARY_DIR}/predict_pb2.py
+      ${CMAKE_CURRENT_BINARY_DIR}/server_test/
+    COMMAND ${CMAKE_COMMAND} -E copy
+      ${grpc_py}
+      ${CMAKE_CURRENT_BINARY_DIR}/server_test/)
+
+    add_test(NAME onnxruntime_server_tests
+    COMMAND onnxruntime_server_tests
+    WORKING_DIRECTORY ${ONNXRUNTIME_SERVER_ROOT}/test/testdata>
+    )
diff --git a/server/ci/run.sh b/server/ci/run.sh
new file mode 100755
index 0000000000000..62ef99b8e4a01
--- /dev/null
+++ b/server/ci/run.sh
@@ -0,0 +1,11 @@
+#!/bin/bash
+set -e
+
+cd /build
+cmake -DCMAKE_BUILD_TYPE=Debug /onnxruntime_src
+make -j$(getconf _NPROCESSORS_ONLN)
+cd /onnxruntime_src/test
+/build/onnxruntime_server_tests
+cd /build
+pip3 install grpcio requests protobuf
+python3 server_test/test_main.py /build/onnxruntime_server /build/models/opset8/test_mnist /onnxruntime_src/test/testdata/server /build /build/server_test
\ No newline at end of file
diff --git a/server/cmake/FindProtobufWithTargets.cmake b/server/cmake/FindProtobufWithTargets.cmake
new file mode 100644
index 0000000000000..6ae90023a04b6
--- /dev/null
+++ b/server/cmake/FindProtobufWithTargets.cmake
@@ -0,0 +1,203 @@
+# ~~~
+# Copyright 2019 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ~~~
+
+#[=======================================================================[.rst:
+FindProtobufWithTargets
+-------------------
+
+A module to use ``Protobuf`` with less complications.
+
+Using ``find_package(Protobuf)`` should be simple, but it is not.
+
+CMake provides a ``FindProtobuf`` module. Unfortunately it does not generate
+``protobuf::*`` targets until CMake-3.9, and ``protobuf::protoc`` does not
+appear until CMake-3.10.
+
+The CMake-config files generated by ``protobuf`` always create these targets,
+but on some Linux distributions (e.g. Fedora>=29, and openSUSE-Tumbleweed) there
+are system packages for protobuf, but these packages are installed without the
+CMake-config files. One must either use the ``FindProtobuf`` module, find the
+libraries via ``pkg-config``, or find the libraries manually.
+
+When the CMake-config files are installed they produce the same targets as
+recent versions of ``FindProtobuf``. However, they do not produce the
+``Protobuf_LIBRARY``, ``Protobuf_INCLUDE_DIR``, etc. that are generated by the
+module. Furthermore, the ``protobuf::protoc`` library is not usable when loaded
+from the CMake-config files: its ``IMPORTED_LOCATION`` variable is not defined.
+
+This module is designed to provide a single, uniform, ``find_package()``
+module that always produces the same outputs:
+
+- It always generates the ``protobuf::*`` targets.
+- It always defines ``ProtobufWithTargets_FOUND`` and
+  ``ProtobufWithTargets_VERSION``.
+- It *prefers* using the CMake config files if they are available.
+- It fallsback on the ``FindProtobuf`` module if the config files are not found.
+- It populates any missing targets and their properties.
+
+The following ``IMPORTED`` targets are defined:
+
+``protobuf::libprotobuf``
+  The protobuf library.
+``protobuf::libprotobuf-lite``
+  The protobuf lite library.
+``protobuf::libprotoc``
+  The protoc library.
+``protobuf::protoc``
+  The protoc compiler.
+
+Example:
+
+.. code-block:: cmake
+
+  find_package(ProtobufWithTargets REQUIRED)
+  add_executable(bar bar.cc)
+  target_link_libraries(bar PRIVATE protobuf::libprotobuf)
+
+#]=======================================================================]
+
+if (protobuf_DEBUG)
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   "protobuf_USE_STATIC_LIBS = ${protobuf_USE_STATIC_LIBS}"
+                   " ProtobufWithTargets = ${ProtobufWithTargets_FOUND}")
+endif ()
+
+# Always load thread support, even on Windows.
+find_package(Threads REQUIRED)
+
+# First try to use the ``protobufConfig.cmake`` or ``protobuf-config.cmake``
+# file if it was installed. This is common on systems (or package managers)
+# where protobuf was compiled and installed with `CMake`. Note that on Linux
+# this *must* be all lowercase ``protobuf``, while on Windows it does not
+# matter.
+find_package(protobuf CONFIG QUIET)
+
+if (protobuf_DEBUG)
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   "protobuf_FOUND = ${protobuf_FOUND}"
+                   " protobuf_VERSION = ${protobuf_VERSION}")
+endif ()
+
+if (NOT protobuf_FOUND)
+    find_package(Protobuf QUIET)
+endif ()
+
+if (protobuf_FOUND)
+    set(ProtobufWithTargets_FOUND 1)
+    set(ProtobufWithTargets_VERSION ${protobuf_VERSION})
+
+    if (NOT TARGET protobuf::libprotobuf)
+        add_library(protobuf::libprotobuf INTERFACE IMPORTED)
+        set_property(
+            TARGET protobuf::libprotobuf PROPERTY INTERFACE_INCLUDE_DIRECTORIES
+                                                  ${Protobuf_INCLUDE_DIR})
+        set_property(
+            TARGET protobuf::libprotobuf
+            APPEND
+            PROPERTY INTERFACE_LINK_LIBRARIES ${Protobuf_LIBRARY}
+                     Threads::Threads)
+    endif ()
+
+    if (NOT TARGET protobuf::libprotobuf-lite)
+        add_library(protobuf::libprotobuf-lite INTERFACE IMPORTED)
+        set_property(
+            TARGET protobuf::libprotobuf-lite
+            PROPERTY INTERFACE_INCLUDE_DIRECTORIES ${Protobuf_INCLUDE_DIR})
+        set_property(
+            TARGET protobuf::libprotobuf-lite
+            APPEND
+            PROPERTY INTERFACE_LINK_LIBRARIES ${Protobuf_LITE_LIBRARY}
+                     Threads::Threads)
+    endif ()
+
+    if (NOT TARGET protobuf::libprotoc)
+        add_library(protobuf::libprotoc INTERFACE IMPORTED)
+        set_property(
+            TARGET protobuf::libprotoc PROPERTY INTERFACE_INCLUDE_DIRECTORIES
+                                                ${Protobuf_INCLUDE_DIR})
+        set_property(
+            TARGET protobuf::libprotoc
+            APPEND
+            PROPERTY INTERFACE_LINK_LIBRARIES ${Protobuf_PROTOC_LIBRARY}
+                     Threads::Threads)
+    endif ()
+
+    if (NOT TARGET protobuf::protoc)
+        add_executable(protobuf::protoc IMPORTED)
+
+        # Discover the protoc compiler location.
+        find_program(
+            _protobuf_PROTOC_EXECUTABLE
+            NAMES protoc
+            DOC "The Google Protocol Buffers Compiler")
+        if (protobuf_DEBUG)
+            message(
+                STATUS
+                    "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                    "ProtobufWithTargets_FOUND = ${ProtobufWithTargets_FOUND}"
+                    " ProtobufWithTargets_VERSION = ${ProtobufWithTargets_VERSION}"
+                    " EXE = ${_protobuf_PROTOC_EXECUTABLE}")
+        endif ()
+        set_property(TARGET protobuf::protoc
+                     PROPERTY IMPORTED_LOCATION ${_protobuf_PROTOC_EXECUTABLE})
+        set_property(
+            TARGET protobuf::protoc PROPERTY IMPORTED_LOCATION_DEBUG
+                                             ${_protobuf_PROTOC_EXECUTABLE})
+        set_property(
+            TARGET protobuf::protoc PROPERTY IMPORTED_LOCATION_RELEASE
+                                             ${_protobuf_PROTOC_EXECUTABLE})
+        unset(_protobuf_PROTOC_EXECUTABLE)
+
+        if (protobuf_DEBUG)
+            get_target_property(_protobuf_PROTOC_EXECUTABLE protobuf::protoc
+                                IMPORTED_LOCATION)
+            message(
+                STATUS
+                    "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                    "LOCATION=${_protobuf_PROTOC_EXECUTABLE}")
+        endif ()
+    endif ()
+endif ()
+
+if (protobuf_DEBUG)
+    message(
+        STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+               "ProtobufWithTargets_FOUND = ${ProtobufWithTargets_FOUND}"
+               " ProtobufWithTargets_VERSION = ${ProtobufWithTargets_VERSION}")
+endif ()
+
+if (protobuf_DEBUG)
+    message(
+        STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+               "ProtobufWithTargets_FOUND = ${ProtobufWithTargets_FOUND}"
+               " ProtobufWithTargets_VERSION = ${ProtobufWithTargets_VERSION}")
+    if (ProtobufWithTargets_FOUND)
+        foreach (_target protobuf::libprotobuf protobuf::libprotobuf-lite
+                 protobuf::libprotoc)
+            if (NOT TARGET ${_target})
+                message(
+                    STATUS
+                        "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                        "target=${_target} is NOT a target")
+            endif ()
+        endforeach ()
+        unset(_target)
+    endif ()
+endif ()
+
+find_package_handle_standard_args(
+    ProtobufWithTargets REQUIRED_VARS ProtobufWithTargets_FOUND
+    ProtobufWithTargets_VERSION)
diff --git a/server/cmake/FindgRPC.cmake b/server/cmake/FindgRPC.cmake
new file mode 100644
index 0000000000000..e6686596a74ca
--- /dev/null
+++ b/server/cmake/FindgRPC.cmake
@@ -0,0 +1,353 @@
+# ~~~
+# Copyright 2019 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ~~~
+
+#[=======================================================================[.rst:
+FindgRPC
+--------
+
+Locate and configure the ``gRPC`` library.
+
+The following variables can be set and are optional:
+
+``gRPC_DEBUG``
+  Show debug messages.
+``gRPC_USE_STATIC_LIBS``
+  Set to ON to force the use of the static libraries.
+  Default is OFF.
+
+Defines the following variables:
+
+``gRPC_FOUND``
+  Found the gRPC library
+``gRPC_VERSION``
+  Version of package found.
+
+The following ``IMPORTED`` targets are also defined:
+
+``gRPC::grpc++``
+  The gRPC C++ library.
+``gRPC::grpc``
+  The gRPC C core library.
+``gRPC::cpp_plugin``
+  The C++ plugin for the Protobuf protoc compiler.
+
+The following cache variables are also available to set or use:
+
+Example:
+
+.. code-block:: cmake
+
+  find_package(gRPC REQUIRED)
+  add_executable(bar bar.cc)
+  target_link_libraries(bar PRIVATE gRPC::grpc++)
+
+#]=======================================================================]
+
+if (gRPC_DEBUG)
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   "gRPC_USE_STATIC_LIBS = ${gRPC_USE_STATIC_LIBS}"
+                   " gRPC_FOUND = ${gRPC_FOUND}")
+endif ()
+
+# gRPC always requires Thread support.
+find_package(Threads REQUIRED)
+
+# Load the module to find protobuf with proper targets. Do not use
+# `find_package()` because we (have to) install this module in non-standard
+# locations.
+include(${CMAKE_CURRENT_LIST_DIR}/FindProtobufWithTargets.cmake)
+
+# The gRPC::grpc_cpp_plugin target is sometimes defined, but without a
+# IMPORTED_LOCATION
+function (_grpc_fix_grpc_cpp_plugin_target)
+    # The target may already exist, do not create it again if it does.
+    if (NOT TARGET gRPC::grpc_cpp_plugin)
+        add_executable(gRPC::grpc_cpp_plugin IMPORTED)
+    endif ()
+    get_target_property(_gRPC_CPP_PLUGIN_EXECUTABLE gRPC::grpc_cpp_plugin
+                        IMPORTED_LOCATION)
+    if (gRPC_DEBUG)
+        message(
+            STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   "LOCATION=${_gRPC_CPP_PLUGIN_EXECUTABLE}")
+    endif ()
+    # Even if the target exists, gRPC CMake support files do not define the
+    # executable for the imported target (at least they do not in v1.19.1), so
+    # we need to define it ourselves.
+    if (NOT _gRPC_CPP_PLUGIN_EXECUTABLE)
+        find_program(_gRPC_CPP_PLUGIN_EXECUTABLE grpc_cpp_plugin
+                     DOC "The gRPC C++ plugin for protoc")
+        mark_as_advanced(_gRPC_CPP_PLUGIN_EXECUTABLE)
+        if (_gRPC_CPP_PLUGIN_EXECUTABLE)
+            set_property(
+                TARGET gRPC::grpc_cpp_plugin
+                PROPERTY IMPORTED_LOCATION ${_gRPC_CPP_PLUGIN_EXECUTABLE})
+        else ()
+            set(gRPC_FOUND "grpc_cpp_plugin-NOTFOUND")
+        endif ()
+    endif ()
+endfunction ()
+
+# The gRPC::* targets sometimes lack the right definitions to compile cleanly on
+# WIN32
+function (_grpc_fix_grpc_target_definitions)
+    # Including gRPC headers without this definition results in a build error.
+    if (WIN32)
+        set_property(
+            TARGET gRPC::grpc
+            APPEND
+            PROPERTY INTERFACE_COMPILE_DEFINITIONS _WIN32_WINNT=0x600)
+        set_property(
+            TARGET gRPC::grpc++
+            APPEND
+            PROPERTY INTERFACE_COMPILE_DEFINITIONS _WIN32_WINNT=0x600)
+    endif ()
+endfunction ()
+
+# First try to use the `gRPCConfig.cmake` or `grpc-config.cmake` file if it was
+# installed. This is common on systems (or package managers) where gRPC was
+# compiled and installed with `CMake`.
+find_package(gRPC NO_MODULE QUIET)
+
+if (gRPC_DEBUG)
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   "NO_MODULE result gRPC_FOUND = ${gRPC_FOUND}")
+endif ()
+
+if (gRPC_FOUND)
+    _grpc_fix_grpc_cpp_plugin_target()
+    _grpc_fix_grpc_target_definitions()
+    return()
+endif ()
+
+include(SelectLibraryConfigurations)
+
+# Internal function: search for normal library as well as a debug one if the
+# debug one is specified also include debug/optimized keywords in *_LIBRARIES
+# variable
+function (_gRPC_find_library name filename)
+    if (${name}_LIBRARY)
+        # Use result recorded by a previous call.
+        return()
+    else ()
+        find_library(${name}_LIBRARY_RELEASE NAMES ${filename})
+        mark_as_advanced(${name}_LIBRARY_RELEASE)
+
+        find_library(${name}_LIBRARY_DEBUG NAMES ${filename}d ${filename})
+        mark_as_advanced(${name}_LIBRARY_DEBUG)
+
+        select_library_configurations(${name})
+
+        if (gRPC_DEBUG)
+            message(
+                STATUS
+                    "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                    "${name} ${filename} RELEASE=${${name}_LIBRARY}"
+                    " DEBUG=${${name}_LIBRARY_DEBUG} DEFAULT=${${name}_LIBRARY}"
+            )
+        endif ()
+
+        set(${name}_LIBRARY
+            "${${name}_LIBRARY}"
+            PARENT_SCOPE)
+    endif ()
+endfunction ()
+
+#
+# Main
+#
+
+# Support preference of static libs by adjusting CMAKE_FIND_LIBRARY_SUFFIXES
+if (_gRPC_USE_STATIC_LIBS)
+    set(_gRPC_ORIG_CMAKE_FIND_LIBRARY_SUFFIXES ${CMAKE_FIND_LIBRARY_SUFFIXES})
+    if (WIN32)
+        set(CMAKE_FIND_LIBRARY_SUFFIXES .lib .a ${CMAKE_FIND_LIBRARY_SUFFIXES})
+    else ()
+        set(CMAKE_FIND_LIBRARY_SUFFIXES .a)
+    endif ()
+endif ()
+
+_grpc_find_library(_gRPC_grpc grpc)
+_grpc_find_library(_gRPC_grpc++ grpc++)
+
+if (NOT _gRPC_INCLUDE_DIR)
+    find_path(_gRPC_INCLUDE_DIR grpcpp/grpcpp.h)
+    mark_as_advanced(_gRPC_INCLUDE_DIR)
+endif ()
+
+if (gRPC_DEBUG)
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   " _gRPC_grpc_LIBRARY = ${_gRPC_grpc_LIBRARY}")
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   " _gRPC_grpc++_LIBRARY = ${_gRPC_grpc++_LIBRARY}")
+    message(STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   " _gRPC_INCLUDE_DIR = ${_gRPC_INCLUDE_DIR}")
+endif ()
+
+if (_gRPC_grpc_LIBRARY)
+    if (NOT TARGET gRPC::grpc)
+        add_library(gRPC::grpc UNKNOWN IMPORTED)
+        set_target_properties(
+            gRPC::grpc PROPERTIES INTERFACE_INCLUDE_DIRECTORIES
+                                  "${_gRPC_INCLUDE_DIR}")
+        if (EXISTS "${_gRPC_grpc_LIBRARY}")
+            set_target_properties(gRPC::grpc PROPERTIES IMPORTED_LOCATION
+                                                        "${_gRPC_grpc_LIBRARY}")
+        endif ()
+        if (EXISTS "${_gRPC_grpc_LIBRARY_RELEASE}")
+            set_property(
+                TARGET gRPC::grpc
+                APPEND
+                PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+            set_target_properties(
+                gRPC::grpc PROPERTIES IMPORTED_LOCATION_RELEASE
+                                      "${_gRPC_grpc_LIBRARY_RELEASE}")
+        endif ()
+        if (EXISTS "${_gRPC_grpc_LIBRARY_DEBUG}")
+            set_property(
+                TARGET gRPC::grpc
+                APPEND
+                PROPERTY IMPORTED_CONFIGURATIONS DEBUG)
+            set_target_properties(
+                gRPC::grpc PROPERTIES IMPORTED_LOCATION_DEBUG
+                                      "${_gRPC_grpc_LIBRARY_DEBUG}")
+        endif ()
+        set_property(
+            TARGET gRPC::grpc
+            APPEND
+            PROPERTY INTERFACE_LINK_LIBRARIES protobuf::libprotobuf
+                     Threads::Threads)
+    endif ()
+endif ()
+
+if (_gRPC_grpc++_LIBRARY)
+    if (NOT TARGET gRPC::grpc++)
+        add_library(gRPC::grpc++ UNKNOWN IMPORTED)
+        set_target_properties(
+            gRPC::grpc++ PROPERTIES INTERFACE_INCLUDE_DIRECTORIES
+                                    "${_gRPC++_INCLUDE_DIR}")
+        if (EXISTS "${_gRPC_grpc++_LIBRARY}")
+            set_target_properties(
+                gRPC::grpc++ PROPERTIES IMPORTED_LOCATION
+                                        "${_gRPC_grpc++_LIBRARY}")
+        endif ()
+        if (EXISTS "${_gRPC_grpc++_LIBRARY_RELEASE}")
+            set_property(
+                TARGET gRPC::grpc++
+                APPEND
+                PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+            set_target_properties(
+                gRPC::grpc++ PROPERTIES IMPORTED_LOCATION_RELEASE
+                                        "${_gRPC_grpc++_LIBRARY_RELEASE}")
+        endif ()
+        if (EXISTS "${_gRPC_grpc++_LIBRARY_DEBUG}")
+            set_property(
+                TARGET gRPC::grpc++
+                APPEND
+                PROPERTY IMPORTED_CONFIGURATIONS DEBUG)
+            set_target_properties(
+                gRPC::grpc++ PROPERTIES IMPORTED_LOCATION_DEBUG
+                                        "${_gRPC_grpc++_LIBRARY_DEBUG}")
+        endif ()
+        set_property(
+            TARGET gRPC::grpc++
+            APPEND
+            PROPERTY INTERFACE_LINK_LIBRARIES gRPC::grpc protobuf::libprotobuf
+                     Threads::Threads)
+        if (CMAKE_VERSION VERSION_GREATER 3.8)
+            # gRPC++ requires C++11, but only CMake-3.8 introduced a target
+            # compiler feature to meet that requirement.
+            set_property(
+                TARGET gRPC::grpc++
+                APPEND
+                PROPERTY INTERFACE_COMPILE_FEATURES cxx_std_11)
+        elseif ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
+            # CMake 3.5 is still alive and kicking in some older distros, use
+            # the compiler-specific versions in these cases.
+            set_property(
+                TARGET gRPC::grpc++
+                APPEND
+                PROPERTY INTERFACE_COMPILE_OPTIONS "-std=c++11")
+        elseif ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
+            set_property(
+                TARGET gRPC::grpc++
+                APPEND
+                PROPERTY INTERFACE_COMPILE_OPTIONS "-std=c++11")
+        else ()
+            message(
+                WARNING
+                    "gRPC::grpc++ requires C++11, but this module"
+                    " (${CMAKE_CURRENT_LIST_FILE})"
+                    " cannot enable it for the library target in your CMake and"
+                    " compiler versions. You need to enable C++11 in the"
+                    " CMakeLists.txt for your project. Consider filing a bug"
+                    " so we can fix this problem.")
+        endif ()
+    endif ()
+endif ()
+
+# Restore original find library prefixes
+if (_gRPC_USE_STATIC_LIBS)
+    set(CMAKE_FIND_LIBRARY_PREFIXES "${_gRPC_ORIG_FIND_LIBRARY_PREFIXES}")
+endif ()
+
+file(
+    WRITE "${CMAKE_BINARY_DIR}/get_gRPC_version.cc"
+    [====[
+#include <grpcpp/grpcpp.h>
+#include <iostream>
+int main() {
+  std::cout << grpc::Version(); // no newline to simplify CMake module
+  return 0;
+}
+        ]====])
+
+try_run(
+    _gRPC_GET_VERSION_STATUS
+    _gRPC_GET_VERSION_COMPILE_STATUS
+    "${CMAKE_BINARY_DIR}"
+    "${CMAKE_BINARY_DIR}/get_gRPC_version.cc"
+    LINK_LIBRARIES
+    gRPC::grpc++
+    gRPC::grpc
+    COMPILE_OUTPUT_VARIABLE _gRPC_GET_VERSION_COMPILE_OUTPUT
+    RUN_OUTPUT_VARIABLE gRPC_VERSION)
+
+file(REMOVE "${CMAKE_BINARY_DIR}/get_gRPC_version.cc")
+
+_grpc_fix_grpc_cpp_plugin_target()
+
+if (gRPC_DEBUG)
+    foreach (
+        _var
+        _gRPC_CPP_PLUGIN_EXECUTABLE
+        _gRPC_VERSION_RAW
+        _gRPC_GET_VERSION_STATUS
+        _gRPC_GET_VERSION_COMPILE_STATUS
+        _gRPC_GET_VERSION_COMPILE_OUTPUT
+        _gRPC_grpc_LIBRARY
+        _gRPC_grpc++_LIBRARY
+        _gRPC_INCLUDE_DIR)
+        message(
+            STATUS "[ ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE} ] "
+                   "${_var} = ${${_var}}")
+    endforeach ()
+    unset(_var)
+endif ()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(gRPC REQUIRED_VARS _gRPC_grpc_LIBRARY
+                                  _gRPC_INCLUDE_DIR VERSION_VAR gRPC_VERSION)
diff --git a/onnxruntime/server/converter.cc b/server/converter.cc
similarity index 88%
rename from onnxruntime/server/converter.cc
rename to server/converter.cc
index 523c022de7545..03ee5d2e42ce1 100644
--- a/onnxruntime/server/converter.cc
+++ b/server/converter.cc
@@ -1,7 +1,7 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 #include "onnx-ml.pb.h"
 #include "predict.pb.h"
@@ -159,35 +159,6 @@ void MLValueToTensorProto(Ort::Value& ml_value, bool using_raw_data,
       }
       break;
     }
-    case onnx::TensorProto_DataType_FLOAT16: {  // Target: raw_data or int32_data
-      const auto* data = ml_value.GetTensorMutableData<onnxruntime::MLFloat16>();
-      if (using_raw_data) {
-        tensor_proto.set_raw_data(data, sizeof(onnxruntime::MLFloat16) * elem_count);
-      } else {
-        for (size_t i = 0, count = elem_count; i < count; ++i) {
-          tensor_proto.add_int32_data(reinterpret_cast<const uint16_t*>(data)[i]);
-        }
-      }
-      break;
-    }
-    case onnx::TensorProto_DataType_BFLOAT16: {  // Target: raw_data or int32_data
-      const auto* data = ml_value.GetTensorMutableData<onnxruntime::BFloat16>();
-
-      std::vector<uint16_t> raw_data;
-      raw_data.reserve(elem_count);
-      for (size_t i = 0; i < elem_count; ++i) {
-        raw_data.push_back(data[i].val);
-      }
-
-      if (using_raw_data) {
-        tensor_proto.set_raw_data(raw_data.data(), raw_data.size() * sizeof(uint16_t));
-      } else {
-        for (size_t i = 0, count = elem_count; i < count; ++i) {
-          tensor_proto.add_int32_data(raw_data[i]);
-        }
-      }
-      break;
-    }
     case onnx::TensorProto_DataType_STRING: {  // Target: string_data
       // string could not be written into "raw_data"
       auto length = ml_value.GetStringTensorDataLength();
@@ -260,4 +231,4 @@ void MLValueToTensorProto(Ort::Value& ml_value, bool using_raw_data,
   return;
 }
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/server/converter.h b/server/converter.h
similarity index 88%
rename from onnxruntime/server/converter.h
rename to server/converter.h
index 3e705f5766204..fc8ef84446526 100644
--- a/onnxruntime/server/converter.h
+++ b/server/converter.h
@@ -2,11 +2,7 @@
 // Licensed under the MIT License.
 
 #pragma once
-#include "core/session/onnxruntime_cxx_api.h"
-
-#include <google/protobuf/stubs/status.h>
-
-#include "core/framework/data_types.h"
+#include "onnxruntime_cxx_api.h"
 
 #include "environment.h"
 #include "predict.pb.h"
diff --git a/onnxruntime/server/core/request_id.cc b/server/core/request_id.cc
similarity index 87%
rename from onnxruntime/server/core/request_id.cc
rename to server/core/request_id.cc
index 8e0c5495d6879..4e08dc9e1fecb 100644
--- a/onnxruntime/server/core/request_id.cc
+++ b/server/core/request_id.cc
@@ -1,3 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
 #include "request_id.h"
 // boost random is using a deprecated header in 1.69
 // See: https://github.com/boostorg/random/issues/49
diff --git a/onnxruntime/server/core/request_id.h b/server/core/request_id.h
similarity index 75%
rename from onnxruntime/server/core/request_id.h
rename to server/core/request_id.h
index b45ca7113ba1a..03e0c9beecf0a 100644
--- a/onnxruntime/server/core/request_id.h
+++ b/server/core/request_id.h
@@ -1,3 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
 #pragma once
 
 #include <string>
diff --git a/onnxruntime/server/environment.cc b/server/environment.cc
similarity index 99%
rename from onnxruntime/server/environment.cc
rename to server/environment.cc
index f928216a380c1..be12314722d6b 100644
--- a/onnxruntime/server/environment.cc
+++ b/server/environment.cc
@@ -3,7 +3,7 @@
 
 #include <memory>
 #include "environment.h"
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 #ifdef USE_DNNL
 
diff --git a/onnxruntime/server/environment.h b/server/environment.h
similarity index 97%
rename from onnxruntime/server/environment.h
rename to server/environment.h
index 86af21029031c..062e4025cd437 100644
--- a/onnxruntime/server/environment.h
+++ b/server/environment.h
@@ -6,7 +6,7 @@
 #include <memory>
 #include <vector>
 
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 #include <spdlog/spdlog.h>
 #include <unordered_map>
 #include <boost/functional/hash.hpp>
diff --git a/onnxruntime/server/executor.cc b/server/executor.cc
similarity index 96%
rename from onnxruntime/server/executor.cc
rename to server/executor.cc
index 52836c844e765..25849d73d313f 100644
--- a/onnxruntime/server/executor.cc
+++ b/server/executor.cc
@@ -2,13 +2,7 @@
 // Licensed under the MIT License.
 
 #include <stdio.h>
-#include "core/common/logging/logging.h"
-#include "core/framework/data_types.h"
-#include "core/session/environment.h"
-#include "core/framework/framework_common.h"
 #include "serializing/mem_buffer.h"
-#include "core/framework/ml_value.h"
-#include "core/framework/tensor.h"
 #include "serializing/tensorprotoutils.h"
 
 #include "onnx-ml.pb.h"
diff --git a/onnxruntime/server/executor.h b/server/executor.h
similarity index 97%
rename from onnxruntime/server/executor.h
rename to server/executor.h
index fbfcd25b20854..7cd7cc3e3dc01 100644
--- a/onnxruntime/server/executor.h
+++ b/server/executor.h
@@ -8,7 +8,7 @@
 #include "environment.h"
 #include "predict.pb.h"
 #include "util.h"
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 namespace onnxruntime {
 namespace server {
diff --git a/cmake/external/spdlog b/server/external/spdlog
similarity index 100%
rename from cmake/external/spdlog
rename to server/external/spdlog
diff --git a/cmake/get_boost.cmake b/server/get_boost.cmake
similarity index 100%
rename from cmake/get_boost.cmake
rename to server/get_boost.cmake
diff --git a/onnxruntime/server/grpc/grpc_app.cc b/server/grpc/grpc_app.cc
similarity index 91%
rename from onnxruntime/server/grpc/grpc_app.cc
rename to server/grpc/grpc_app.cc
index fe73b91abdc4c..1d571b43fe012 100644
--- a/onnxruntime/server/grpc/grpc_app.cc
+++ b/server/grpc/grpc_app.cc
@@ -1,3 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
 #include "grpc_app.h"
 #include <grpcpp/health_check_service_interface.h>
 #include <grpcpp/ext/channelz_service_plugin.h>
diff --git a/onnxruntime/server/grpc/grpc_app.h b/server/grpc/grpc_app.h
similarity index 87%
rename from onnxruntime/server/grpc/grpc_app.h
rename to server/grpc/grpc_app.h
index 27cea08d15155..a9345c34441ff 100644
--- a/onnxruntime/server/grpc/grpc_app.h
+++ b/server/grpc/grpc_app.h
@@ -1,3 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
 #pragma once
 #include <grpcpp/grpcpp.h>
 #include "prediction_service_impl.h"
diff --git a/onnxruntime/server/grpc/prediction_service_impl.cc b/server/grpc/prediction_service_impl.cc
similarity index 94%
rename from onnxruntime/server/grpc/prediction_service_impl.cc
rename to server/grpc/prediction_service_impl.cc
index cd2817abab300..834f5c894939a 100644
--- a/onnxruntime/server/grpc/prediction_service_impl.cc
+++ b/server/grpc/prediction_service_impl.cc
@@ -1,3 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
 #include "prediction_service_impl.h"
 #include "request_id.h"
 
diff --git a/onnxruntime/server/grpc/prediction_service_impl.h b/server/grpc/prediction_service_impl.h
similarity index 89%
rename from onnxruntime/server/grpc/prediction_service_impl.h
rename to server/grpc/prediction_service_impl.h
index b1024d31c28c2..be8d7ecb694f3 100644
--- a/onnxruntime/server/grpc/prediction_service_impl.h
+++ b/server/grpc/prediction_service_impl.h
@@ -1,3 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
 #pragma once
 #include "prediction_service.grpc.pb.h"
 #include "environment.h"
diff --git a/onnxruntime/server/http/core/context.h b/server/http/core/context.h
similarity index 100%
rename from onnxruntime/server/http/core/context.h
rename to server/http/core/context.h
diff --git a/onnxruntime/server/http/core/http_server.cc b/server/http/core/http_server.cc
similarity index 100%
rename from onnxruntime/server/http/core/http_server.cc
rename to server/http/core/http_server.cc
diff --git a/onnxruntime/server/http/core/http_server.h b/server/http/core/http_server.h
similarity index 100%
rename from onnxruntime/server/http/core/http_server.h
rename to server/http/core/http_server.h
diff --git a/onnxruntime/server/http/core/listener.cc b/server/http/core/listener.cc
similarity index 100%
rename from onnxruntime/server/http/core/listener.cc
rename to server/http/core/listener.cc
diff --git a/onnxruntime/server/http/core/listener.h b/server/http/core/listener.h
similarity index 100%
rename from onnxruntime/server/http/core/listener.h
rename to server/http/core/listener.h
diff --git a/onnxruntime/server/http/core/routes.cc b/server/http/core/routes.cc
similarity index 100%
rename from onnxruntime/server/http/core/routes.cc
rename to server/http/core/routes.cc
diff --git a/onnxruntime/server/http/core/routes.h b/server/http/core/routes.h
similarity index 100%
rename from onnxruntime/server/http/core/routes.h
rename to server/http/core/routes.h
diff --git a/onnxruntime/server/http/core/session.cc b/server/http/core/session.cc
similarity index 100%
rename from onnxruntime/server/http/core/session.cc
rename to server/http/core/session.cc
diff --git a/onnxruntime/server/http/core/session.h b/server/http/core/session.h
similarity index 100%
rename from onnxruntime/server/http/core/session.h
rename to server/http/core/session.h
diff --git a/onnxruntime/server/http/core/util.cc b/server/http/core/util.cc
similarity index 100%
rename from onnxruntime/server/http/core/util.cc
rename to server/http/core/util.cc
diff --git a/onnxruntime/server/http/core/util.h b/server/http/core/util.h
similarity index 100%
rename from onnxruntime/server/http/core/util.h
rename to server/http/core/util.h
diff --git a/onnxruntime/server/http/json_handling.cc b/server/http/json_handling.cc
similarity index 100%
rename from onnxruntime/server/http/json_handling.cc
rename to server/http/json_handling.cc
diff --git a/onnxruntime/server/http/json_handling.h b/server/http/json_handling.h
similarity index 100%
rename from onnxruntime/server/http/json_handling.h
rename to server/http/json_handling.h
diff --git a/onnxruntime/server/http/predict_request_handler.cc b/server/http/predict_request_handler.cc
similarity index 100%
rename from onnxruntime/server/http/predict_request_handler.cc
rename to server/http/predict_request_handler.cc
diff --git a/onnxruntime/server/http/predict_request_handler.h b/server/http/predict_request_handler.h
similarity index 100%
rename from onnxruntime/server/http/predict_request_handler.h
rename to server/http/predict_request_handler.h
diff --git a/onnxruntime/server/http/util.cc b/server/http/util.cc
similarity index 100%
rename from onnxruntime/server/http/util.cc
rename to server/http/util.cc
diff --git a/onnxruntime/server/http/util.h b/server/http/util.h
similarity index 100%
rename from onnxruntime/server/http/util.h
rename to server/http/util.h
diff --git a/onnxruntime/server/main.cc b/server/main.cc
similarity index 100%
rename from onnxruntime/server/main.cc
rename to server/main.cc
diff --git a/onnxruntime/server/protobuf/onnx-ml.proto b/server/protobuf/onnx-ml.proto
similarity index 100%
rename from onnxruntime/server/protobuf/onnx-ml.proto
rename to server/protobuf/onnx-ml.proto
diff --git a/onnxruntime/server/protobuf/predict.proto b/server/protobuf/predict.proto
similarity index 100%
rename from onnxruntime/server/protobuf/predict.proto
rename to server/protobuf/predict.proto
diff --git a/onnxruntime/server/protobuf/prediction_service.proto b/server/protobuf/prediction_service.proto
similarity index 100%
rename from onnxruntime/server/protobuf/prediction_service.proto
rename to server/protobuf/prediction_service.proto
diff --git a/server/protobuf_function.cmake b/server/protobuf_function.cmake
new file mode 100644
index 0000000000000..d3c0b6e688c66
--- /dev/null
+++ b/server/protobuf_function.cmake
@@ -0,0 +1,158 @@
+# Protocol Buffers - Google's data interchange format
+# Copyright 2008 Google Inc.  All rights reserved.
+# https://developers.google.com/protocol-buffers/
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are
+# met:
+#
+#     * Redistributions of source code must retain the above copyright
+# notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above
+# copyright notice, this list of conditions and the following disclaimer
+# in the documentation and/or other materials provided with the
+# distribution.
+#     * Neither the name of Google Inc. nor the names of its
+# contributors may be used to endorse or promote products derived from
+# this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#Changelog:
+#copied from https://github.com/protocolbuffers/protobuf/blob/master/cmake/protobuf-config.cmake.in
+#sed -i 's/protobuf_generate/onnxruntime_protobuf_generate/g' protobuf-config.cmake.orig
+#replace 'protobuf::protoc' with ${PROTOC_EXECUTABLE} and ${PROTOC_DEPS}
+#remove OUTDIR
+
+function(onnxruntime_protobuf_generate)
+  include(CMakeParseArguments)
+  if(EXISTS "${ONNX_CUSTOM_PROTOC_EXECUTABLE}")
+    set(PROTOC_EXECUTABLE ${ONNX_CUSTOM_PROTOC_EXECUTABLE})
+  else()
+    set(PROTOC_EXECUTABLE $<TARGET_FILE:protobuf::protoc>)
+    set(PROTOC_DEPS protobuf::protoc)
+  endif()
+  set(_options APPEND_PATH)
+  set(_singleargs LANGUAGE OUT_VAR EXPORT_MACRO)
+  if(COMMAND target_sources)
+    list(APPEND _singleargs TARGET)
+  endif()
+  set(_multiargs PROTOS IMPORT_DIRS GENERATE_EXTENSIONS)
+
+  cmake_parse_arguments(onnxruntime_protobuf_generate "${_options}" "${_singleargs}" "${_multiargs}" "${ARGN}")
+
+  if(NOT onnxruntime_protobuf_generate_PROTOS AND NOT onnxruntime_protobuf_generate_TARGET)
+    message(SEND_ERROR "Error: onnxruntime_protobuf_generate called without any targets or source files")
+    return()
+  endif()
+
+  if(NOT onnxruntime_protobuf_generate_OUT_VAR AND NOT onnxruntime_protobuf_generate_TARGET)
+    message(SEND_ERROR "Error: onnxruntime_protobuf_generate called without a target or output variable")
+    return()
+  endif()
+
+  if(NOT onnxruntime_protobuf_generate_LANGUAGE)
+    set(onnxruntime_protobuf_generate_LANGUAGE cpp)
+  endif()
+  string(TOLOWER ${onnxruntime_protobuf_generate_LANGUAGE} onnxruntime_protobuf_generate_LANGUAGE)
+
+  if(onnxruntime_protobuf_generate_EXPORT_MACRO AND onnxruntime_protobuf_generate_LANGUAGE STREQUAL cpp)
+    set(_dll_export_decl "dllexport_decl=${onnxruntime_protobuf_generate_EXPORT_MACRO}:")
+  endif()
+
+  if(NOT onnxruntime_protobuf_generate_EXTENSIONS)
+    if(onnxruntime_protobuf_generate_LANGUAGE STREQUAL cpp)
+      set(onnxruntime_protobuf_generate_EXTENSIONS .pb.h .pb.cc)
+    elseif(onnxruntime_protobuf_generate_LANGUAGE STREQUAL python)
+      set(onnxruntime_protobuf_generate_EXTENSIONS _pb2.py)
+    else()
+      message(SEND_ERROR "Error: onnxruntime_protobuf_generate given unknown Language ${LANGUAGE}, please provide a value for GENERATE_EXTENSIONS")
+      return()
+    endif()
+  endif()
+
+  if(onnxruntime_protobuf_generate_TARGET)
+    get_target_property(_source_list ${onnxruntime_protobuf_generate_TARGET} SOURCES)
+    foreach(_file ${_source_list})
+      if(_file MATCHES "proto$")
+        list(APPEND onnxruntime_protobuf_generate_PROTOS ${_file})
+      endif()
+    endforeach()
+  endif()
+
+  if(NOT onnxruntime_protobuf_generate_PROTOS)
+    message(SEND_ERROR "Error: onnxruntime_protobuf_generate could not find any .proto files")
+    return()
+  endif()
+
+  if(onnxruntime_protobuf_generate_APPEND_PATH)
+    # Create an include path for each file specified
+    foreach(_file ${onnxruntime_protobuf_generate_PROTOS})
+      get_filename_component(_abs_file ${_file} ABSOLUTE)
+      get_filename_component(_abs_path ${_abs_file} PATH)
+      list(FIND _protobuf_include_path ${_abs_path} _contains_already)
+      if(${_contains_already} EQUAL -1)
+        list(APPEND _protobuf_include_path -I ${_abs_path})
+      endif()
+    endforeach()
+  else()
+    set(_protobuf_include_path -I ${CMAKE_CURRENT_SOURCE_DIR})
+  endif()
+
+  foreach(DIR ${onnxruntime_protobuf_generate_IMPORT_DIRS})
+    get_filename_component(ABS_PATH ${DIR} ABSOLUTE)
+    list(FIND _protobuf_include_path ${ABS_PATH} _contains_already)
+    if(${_contains_already} EQUAL -1)
+      list(APPEND _protobuf_include_path -I ${ABS_PATH})
+    endif()
+  endforeach()
+
+  set(_generated_srcs_all)
+  foreach(_proto ${onnxruntime_protobuf_generate_PROTOS})
+    get_filename_component(_abs_file ${_proto} ABSOLUTE)
+    get_filename_component(_basename ${_proto} NAME_WE)
+
+    set(_generated_srcs)
+    foreach(_ext ${onnxruntime_protobuf_generate_EXTENSIONS})
+      list(APPEND _generated_srcs "${CMAKE_CURRENT_BINARY_DIR}/${_basename}${_ext}")
+    endforeach()
+    list(APPEND _generated_srcs_all ${_generated_srcs})
+
+    if (onnxruntime_USE_FULL_PROTOBUF)
+      add_custom_command(
+        OUTPUT ${_generated_srcs}
+        COMMAND  ${PROTOC_EXECUTABLE}
+        ARGS --${onnxruntime_protobuf_generate_LANGUAGE}_out ${_dll_export_decl}${CMAKE_CURRENT_BINARY_DIR} ${_protobuf_include_path} ${_abs_file}
+        DEPENDS ${_abs_file} ${PROTOC_DEPS}
+        COMMENT "Running ${onnxruntime_protobuf_generate_LANGUAGE} protocol buffer (full) compiler on ${_proto}"
+        VERBATIM )
+    else()
+      add_custom_command(
+        OUTPUT ${_generated_srcs}
+        COMMAND  ${PROTOC_EXECUTABLE}
+        ARGS --${onnxruntime_protobuf_generate_LANGUAGE}_out lite:${_dll_export_decl}${CMAKE_CURRENT_BINARY_DIR} ${_protobuf_include_path} ${_abs_file}
+        DEPENDS ${_abs_file} ${PROTOC_DEPS}
+        COMMENT "Running ${onnxruntime_protobuf_generate_LANGUAGE} protocol buffer compiler (lite) on ${_proto}"
+        VERBATIM )
+    endif()
+  endforeach()
+
+  set_source_files_properties(${_generated_srcs_all} PROPERTIES GENERATED TRUE)
+  if(onnxruntime_protobuf_generate_OUT_VAR)
+    set(${onnxruntime_protobuf_generate_OUT_VAR} ${_generated_srcs_all} PARENT_SCOPE)
+  endif()
+  if(onnxruntime_protobuf_generate_TARGET)
+    target_sources(${onnxruntime_protobuf_generate_TARGET} PRIVATE ${_generated_srcs_all})
+  endif()
+
+endfunction()
diff --git a/onnxruntime/server/serializing/mem_buffer.h b/server/serializing/mem_buffer.h
similarity index 72%
rename from onnxruntime/server/serializing/mem_buffer.h
rename to server/serializing/mem_buffer.h
index bcda8ce0e189f..fa2f28b73a028 100644
--- a/onnxruntime/server/serializing/mem_buffer.h
+++ b/server/serializing/mem_buffer.h
@@ -1,5 +1,9 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
 #pragma once
-#include "core/common/common.h"
+
+#include "onnxruntime_c_api.h"
+#include "onnxruntime_cxx_api.h"
 
 namespace onnxruntime {
 namespace server {
@@ -18,4 +22,4 @@ class MemBuffer {
   const OrtMemoryInfo& alloc_info_;
 };
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/server/serializing/tensorprotoutils.cc b/server/serializing/tensorprotoutils.cc
similarity index 82%
rename from onnxruntime/server/serializing/tensorprotoutils.cc
rename to server/serializing/tensorprotoutils.cc
index 45e062374e790..ee6d6abf7ffae 100644
--- a/onnxruntime/server/serializing/tensorprotoutils.cc
+++ b/server/serializing/tensorprotoutils.cc
@@ -6,24 +6,12 @@
 #include <memory>
 #include <algorithm>
 #include <limits>
-#include <gsl/gsl>
-#include "core/framework/data_types.h"
-#include "core/framework/allocator.h"
+#include <sstream>
 #include "onnx-ml.pb.h"
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 namespace onnxruntime {
-namespace server {
-#ifdef __GNUC__
-constexpr inline bool IsLittleEndianOrder() noexcept { return __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__; }
-#else
-// On Windows and Mac, this function should always return true
-GSL_SUPPRESS(type .1)  // allow use of reinterpret_cast for this special case
-inline bool IsLittleEndianOrder() noexcept {
-  static int n = 1;
-  return (*reinterpret_cast<char*>(&n) == 1);
-}
-#endif
+
 
 //From core common
 inline void MakeStringInternal(std::ostringstream& /*ss*/) noexcept {
@@ -56,6 +44,19 @@ inline std::string MakeString(const char* p_str) {
   return p_str;
 }
 
+
+
+namespace server {
+#ifdef __GNUC__
+constexpr inline bool IsLittleEndianOrder() noexcept { return __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__; }
+#else
+// On Windows and Mac, this function should always return true
+GSL_SUPPRESS(type .1)  // allow use of reinterpret_cast for this special case
+inline bool IsLittleEndianOrder() noexcept {
+  static int n = 1;
+  return (*reinterpret_cast<char*>(&n) == 1);
+}
+#endif
 std::vector<int64_t> GetTensorShapeFromTensorProto(const onnx::TensorProto& tensor_proto) {
   const auto& dims = tensor_proto.dims();
   std::vector<int64_t> tensor_shape_vec(static_cast<size_t>(dims.size()));
@@ -66,17 +67,40 @@ std::vector<int64_t> GetTensorShapeFromTensorProto(const onnx::TensorProto& tens
   return tensor_shape_vec;
 }
 
+
+template <size_t alignment>
+static bool CalcMemSizeForArrayWithAlignment(size_t nmemb, size_t size, size_t* out) noexcept {
+  static constexpr size_t max_allowed = (static_cast<size_t>(1) << (static_cast<size_t>(std::numeric_limits<size_t>::digits >> 1))) - alignment;
+  static constexpr size_t max_size = std::numeric_limits<size_t>::max() - alignment;
+  static constexpr size_t alignment_mask = alignment - 1;
+  //Indeed, we only need to check if max_size / nmemb < size
+  //max_allowed is for avoiding unnecessary DIV.
+  if (nmemb >= max_allowed && max_size / nmemb < size) {
+    return false;
+  }
+  if (size >= max_allowed &&
+      nmemb > 0 && max_size / nmemb < size) {
+    return false;
+  }
+  if (alignment == 0)
+    *out = size * nmemb;
+  else
+    *out = (size * nmemb + alignment_mask) & ~static_cast<size_t>(alignment_mask);
+  return true;
+}
+
+static bool CalcMemSizeForArray(size_t nmemb, size_t size, size_t* out) noexcept {
+  return CalcMemSizeForArrayWithAlignment<0>(nmemb, size, out);
+}
+
 // This function doesn't support string tensors
 template <typename T>
 static void UnpackTensorWithRawData(const void* raw_data, size_t raw_data_length, size_t expected_size,
                                     /*out*/ T* p_data) {
   // allow this low level routine to be somewhat unsafe. assuming it's thoroughly tested and valid
-  GSL_SUPPRESS(type)       // type.1 reinterpret-cast; type.4 C-style casts; type.5 'T result;' is uninitialized;
-  GSL_SUPPRESS(bounds .1)  // pointer arithmetic
-  GSL_SUPPRESS(f .23)      // buff and temp_bytes never tested for nullness and could be gsl::not_null
   {
     size_t expected_size_in_bytes;
-    if (!onnxruntime::IAllocator::CalcMemSizeForArray(expected_size, sizeof(T), &expected_size_in_bytes)) {
+    if (!CalcMemSizeForArray(expected_size, sizeof(T), &expected_size_in_bytes)) {
       throw Ort::Exception("size overflow", OrtErrorCode::ORT_FAIL);
     }
     if (raw_data_length != expected_size_in_bytes)
@@ -187,77 +211,11 @@ void UnpackTensor(const onnx::TensorProto& tensor, const void* raw_data, size_t
 
   return;
 }
-template <>
-void UnpackTensor(const onnx::TensorProto& tensor, const void* raw_data, size_t raw_data_len,
-                  /*out*/ MLFloat16* p_data, int64_t expected_size) {
-  if (nullptr == p_data) {
-    const size_t size = raw_data != nullptr ? raw_data_len : tensor.int32_data_size();
-    if (size == 0) return;
-    throw Ort::Exception("", OrtErrorCode::ORT_INVALID_ARGUMENT);
-  }
-  if (onnx::TensorProto_DataType_FLOAT16 != tensor.data_type()) {
-    throw Ort::Exception("", OrtErrorCode::ORT_INVALID_ARGUMENT);
-  }
-
-  if (raw_data != nullptr) {
-    return UnpackTensorWithRawData(raw_data, raw_data_len, expected_size, p_data);
-  }
-
-  if (tensor.int32_data_size() != expected_size)
-    throw Ort::Exception(
-        "UnpackTensor: the pre-allocate size does not match the size in proto", OrtErrorCode::ORT_FAIL);
 
-  constexpr int max_value = std::numeric_limits<uint16_t>::max();
-  for (int i = 0; i < static_cast<int>(expected_size); i++) {
-    int v = tensor.int32_data()[i];
-    if (v < 0 || v > max_value) {
-      throw Ort::Exception(
-          "data overflow", OrtErrorCode::ORT_FAIL);
-    }
-    p_data[i] = MLFloat16(static_cast<uint16_t>(v));
-  }
-
-  return;
-}
-
-template <>
-void UnpackTensor(const onnx::TensorProto& tensor, const void* raw_data, size_t raw_data_len,
-                  /*out*/ BFloat16* p_data, int64_t expected_size) {
-  if (nullptr == p_data) {
-    const size_t size = raw_data != nullptr ? raw_data_len : tensor.int32_data_size();
-    if (size == 0)
-      return;
-
-    throw Ort::Exception("", OrtErrorCode::ORT_INVALID_ARGUMENT);
-  }
-  if (onnx::TensorProto_DataType_BFLOAT16 != tensor.data_type()) {
-    throw Ort::Exception("", OrtErrorCode::ORT_INVALID_ARGUMENT);
-  }
-
-  if (raw_data != nullptr) {
-    return UnpackTensorWithRawData(raw_data, raw_data_len, expected_size, p_data);
-  }
-
-  if (tensor.int32_data_size() != expected_size)
-    throw Ort::Exception(
-        "UnpackTensor: the pre-allocate size does not match the size in proto", OrtErrorCode::ORT_FAIL);
-
-  constexpr int max_value = std::numeric_limits<uint16_t>::max();
-  for (int i = 0; i < static_cast<int>(expected_size); i++) {
-    int v = tensor.int32_data()[i];
-    if (v < 0 || v > max_value) {
-      throw Ort::Exception(
-          "data overflow", OrtErrorCode::ORT_FAIL);
-    }
-    p_data[i] = BFloat16(static_cast<uint16_t>(v));
-  }
-
-  return;
-}
 
 #define CASE_PROTO_TRACE(X, Y)                                                            \
   case onnx::TensorProto_DataType::TensorProto_DataType_##X:                              \
-    if (!IAllocator::CalcMemSizeForArrayWithAlignment<alignment>(size, sizeof(Y), out)) { \
+    if (!CalcMemSizeForArrayWithAlignment<alignment>(size, sizeof(Y), out)) { \
       throw Ort::Exception("Invalid TensorProto", OrtErrorCode::ORT_FAIL);                \
     }                                                                                     \
     break;
@@ -270,7 +228,7 @@ void GetSizeInBytesFromTensorProto(const onnx::TensorProto& tensor_proto, size_t
     if (dim < 0 || static_cast<uint64_t>(dim) >= std::numeric_limits<size_t>::max()) {
       throw Ort::Exception("Invalid TensorProto", OrtErrorCode::ORT_FAIL);
     }
-    if (!IAllocator::CalcMemSizeForArray(size, static_cast<size_t>(dim), &size)) {
+    if (!CalcMemSizeForArray(size, static_cast<size_t>(dim), &size)) {
       throw Ort::Exception("Invalid TensorProto", OrtErrorCode::ORT_FAIL);
     }
   }
@@ -286,8 +244,6 @@ void GetSizeInBytesFromTensorProto(const onnx::TensorProto& tensor_proto, size_t
     CASE_PROTO_TRACE(UINT16, uint16_t);
     CASE_PROTO_TRACE(UINT32, uint32_t);
     CASE_PROTO_TRACE(UINT64, uint64_t);
-    CASE_PROTO_TRACE(FLOAT16, MLFloat16);
-    CASE_PROTO_TRACE(BFLOAT16, BFloat16);
     CASE_PROTO_TRACE(STRING, std::string);
     default:
       throw Ort::Exception("", OrtErrorCode::ORT_NOT_IMPLEMENTED);
@@ -401,8 +357,6 @@ void TensorProtoToMLValue(const onnx::TensorProto& tensor_proto, const MemBuffer
         CASE_PROTO(UINT16, uint16_t);
         CASE_PROTO(UINT32, uint32_t);
         CASE_PROTO(UINT64, uint64_t);
-        CASE_PROTO(FLOAT16, MLFloat16);
-        CASE_PROTO(BFLOAT16, BFloat16);
         case onnx::TensorProto_DataType::TensorProto_DataType_STRING:
           if (preallocated != nullptr) {
             OrtInitializeBufferForTensor(preallocated, preallocated_size, ele_type);
@@ -428,4 +382,4 @@ template void GetSizeInBytesFromTensorProto<256>(const onnx::TensorProto& tensor
                                                  size_t* out);
 template void GetSizeInBytesFromTensorProto<0>(const onnx::TensorProto& tensor_proto, size_t* out);
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/server/serializing/tensorprotoutils.h b/server/serializing/tensorprotoutils.h
similarity index 91%
rename from onnxruntime/server/serializing/tensorprotoutils.h
rename to server/serializing/tensorprotoutils.h
index 243b122ef64b5..4b08a29448efa 100644
--- a/onnxruntime/server/serializing/tensorprotoutils.h
+++ b/server/serializing/tensorprotoutils.h
@@ -5,8 +5,8 @@
 
 #include <vector>
 #include <type_traits>
-#include "core/session/onnxruntime_c_api.h"
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_c_api.h"
+#include "onnxruntime_cxx_api.h"
 
 #include "mem_buffer.h"
 
@@ -34,4 +34,4 @@ void UnpackTensor(const onnx::TensorProto& tensor, const void* raw_data, size_t
 ONNXTensorElementDataType CApiElementTypeFromProtoType(int type);
 ONNXTensorElementDataType GetTensorElementType(const onnx::TensorProto& tensor_proto);
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/server/server_configuration.h b/server/server_configuration.h
similarity index 99%
rename from onnxruntime/server/server_configuration.h
rename to server/server_configuration.h
index 3fb075a4776f4..eeb883f8999a7 100644
--- a/onnxruntime/server/server_configuration.h
+++ b/server/server_configuration.h
@@ -8,7 +8,7 @@
 #include <unordered_map>
 
 #include "boost/program_options.hpp"
-#include "core/session/onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 namespace onnxruntime {
 namespace server {
diff --git a/onnxruntime/test/server/integration_tests/README.MD b/server/test/integration_tests/README.MD
similarity index 100%
rename from onnxruntime/test/server/integration_tests/README.MD
rename to server/test/integration_tests/README.MD
diff --git a/onnxruntime/test/server/integration_tests/function_tests.py b/server/test/integration_tests/function_tests.py
similarity index 100%
rename from onnxruntime/test/server/integration_tests/function_tests.py
rename to server/test/integration_tests/function_tests.py
diff --git a/onnxruntime/test/server/integration_tests/model_zoo_data_prep.py b/server/test/integration_tests/model_zoo_data_prep.py
similarity index 100%
rename from onnxruntime/test/server/integration_tests/model_zoo_data_prep.py
rename to server/test/integration_tests/model_zoo_data_prep.py
diff --git a/onnxruntime/test/server/integration_tests/model_zoo_tests.py b/server/test/integration_tests/model_zoo_tests.py
similarity index 100%
rename from onnxruntime/test/server/integration_tests/model_zoo_tests.py
rename to server/test/integration_tests/model_zoo_tests.py
diff --git a/onnxruntime/test/server/integration_tests/test_main.py b/server/test/integration_tests/test_main.py
similarity index 100%
rename from onnxruntime/test/server/integration_tests/test_main.py
rename to server/test/integration_tests/test_main.py
diff --git a/onnxruntime/test/server/integration_tests/test_util.py b/server/test/integration_tests/test_util.py
similarity index 100%
rename from onnxruntime/test/server/integration_tests/test_util.py
rename to server/test/integration_tests/test_util.py
diff --git a/server/test/testdata/mul_1.onnx b/server/test/testdata/mul_1.onnx
new file mode 100644
index 0000000000000000000000000000000000000000..0b6dc510261322f3ff63deaa4114a0a023b35246
GIT binary patch
literal 130
zcmd;J7Gg`zNX;urv?|~d<zkHBVhk5zjFe){EzOBHRATlm%@N`NvK2~Fi%YbWIhZ+^
z1Q?Yh7#JGt85kHGfCz{ifOr8AA8>GD43829nj*x<#lyiU#KFbH!3@MqNkG9!bU`N;
IE(QU106-ZNs{jB1

literal 0
HcmV?d00001

diff --git a/onnxruntime/test/testdata/server/mnist_test_data_set_0_input.json b/server/test/testdata/server/mnist_test_data_set_0_input.json
similarity index 100%
rename from onnxruntime/test/testdata/server/mnist_test_data_set_0_input.json
rename to server/test/testdata/server/mnist_test_data_set_0_input.json
diff --git a/onnxruntime/test/testdata/server/mnist_test_data_set_0_input.pb b/server/test/testdata/server/mnist_test_data_set_0_input.pb
similarity index 100%
rename from onnxruntime/test/testdata/server/mnist_test_data_set_0_input.pb
rename to server/test/testdata/server/mnist_test_data_set_0_input.pb
diff --git a/onnxruntime/test/testdata/server/mnist_test_data_set_0_output.json b/server/test/testdata/server/mnist_test_data_set_0_output.json
similarity index 100%
rename from onnxruntime/test/testdata/server/mnist_test_data_set_0_output.json
rename to server/test/testdata/server/mnist_test_data_set_0_output.json
diff --git a/onnxruntime/test/testdata/server/mnist_test_data_set_0_output.pb b/server/test/testdata/server/mnist_test_data_set_0_output.pb
similarity index 100%
rename from onnxruntime/test/testdata/server/mnist_test_data_set_0_output.pb
rename to server/test/testdata/server/mnist_test_data_set_0_output.pb
diff --git a/onnxruntime/test/testdata/server/request_0.pb b/server/test/testdata/server/request_0.pb
similarity index 100%
rename from onnxruntime/test/testdata/server/request_0.pb
rename to server/test/testdata/server/request_0.pb
diff --git a/onnxruntime/test/testdata/server/response_0.pb b/server/test/testdata/server/response_0.pb
similarity index 100%
rename from onnxruntime/test/testdata/server/response_0.pb
rename to server/test/testdata/server/response_0.pb
diff --git a/onnxruntime/test/server/unit_tests/executor_test.cc b/server/test/unit_tests/executor_test.cc
similarity index 96%
rename from onnxruntime/test/server/unit_tests/executor_test.cc
rename to server/test/unit_tests/executor_test.cc
index 40561198d8a1d..0ff196f63a127 100644
--- a/onnxruntime/test/server/unit_tests/executor_test.cc
+++ b/server/test/unit_tests/executor_test.cc
@@ -5,8 +5,8 @@
 
 #include "gtest/gtest.h"
 
-#include "server/executor.h"
-#include "server/http/json_handling.h"
+#include "executor.h"
+#include "http/json_handling.h"
 #include <spdlog/spdlog.h>
 #include <spdlog/sinks/sink.h>
 #include <spdlog/sinks/stdout_sinks.h>
diff --git a/onnxruntime/test/server/unit_tests/external/server_context_test_spouse.h b/server/test/unit_tests/external/server_context_test_spouse.h
similarity index 100%
rename from onnxruntime/test/server/unit_tests/external/server_context_test_spouse.h
rename to server/test/unit_tests/external/server_context_test_spouse.h
diff --git a/onnxruntime/test/server/unit_tests/http_routes_tests.cc b/server/test/unit_tests/http_routes_tests.cc
similarity index 99%
rename from onnxruntime/test/server/unit_tests/http_routes_tests.cc
rename to server/test/unit_tests/http_routes_tests.cc
index 842eabb071134..e084a8befd907 100644
--- a/onnxruntime/test/server/unit_tests/http_routes_tests.cc
+++ b/server/test/unit_tests/http_routes_tests.cc
@@ -4,7 +4,7 @@
 #include <iostream>
 
 #include "gtest/gtest.h"
-#include "server/http/core/routes.h"
+#include "http/core/routes.h"
 
 namespace onnxruntime {
 namespace server {
diff --git a/onnxruntime/test/server/unit_tests/json_handling_tests.cc b/server/test/unit_tests/json_handling_tests.cc
similarity index 98%
rename from onnxruntime/test/server/unit_tests/json_handling_tests.cc
rename to server/test/unit_tests/json_handling_tests.cc
index 0bd2d16e45795..5e6f8dfc0a416 100644
--- a/onnxruntime/test/server/unit_tests/json_handling_tests.cc
+++ b/server/test/unit_tests/json_handling_tests.cc
@@ -7,7 +7,7 @@
 #include "gtest/gtest.h"
 
 #include "predict.pb.h"
-#include "server/http/json_handling.h"
+#include "http/json_handling.h"
 
 namespace onnxruntime {
 namespace server {
@@ -36,7 +36,7 @@ TEST(JsonDeserializationTests, InvalidData) {
   protobufutil::Status status = onnxruntime::server::GetRequestFromJson(input_json, request);
 
   EXPECT_EQ(protobufutil::error::INVALID_ARGUMENT, status.error_code());
-  EXPECT_EQ("inputs[0].value.raw_data: invalid value \"hello\" for type TYPE_BYTES", status.error_message());
+  EXPECT_EQ("(inputs[0].value.raw_data): invalid value \"hello\" for type TYPE_BYTES", status.error_message()) << status.error_message();
 }
 
 TEST(JsonDeserializationTests, InvalidJson) {
@@ -125,4 +125,4 @@ TEST(JsonErrorMessageTests, MessageWithManyCarriageCharacters) {
 
 }  // namespace test
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/test/server/unit_tests/prediction_service_impl_test.cc b/server/test/unit_tests/prediction_service_impl_test.cc
similarity index 96%
rename from onnxruntime/test/server/unit_tests/prediction_service_impl_test.cc
rename to server/test/unit_tests/prediction_service_impl_test.cc
index 39f32b62c1ad8..968a42c1512d5 100644
--- a/onnxruntime/test/server/unit_tests/prediction_service_impl_test.cc
+++ b/server/test/unit_tests/prediction_service_impl_test.cc
@@ -1,10 +1,11 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
 #include <iostream>
 
 #include "gtest/gtest.h"
 
-#include "server/executor.h"
-#include "server/grpc/prediction_service_impl.h"
-#include "test/test_environment.h"
+#include "executor.h"
+#include "grpc/prediction_service_impl.h"
 #include "test_server_environment.h"
 #include "external/server_context_test_spouse.h"
 #include <grpcpp/impl/grpc_library.h>
diff --git a/onnxruntime/test/server/unit_tests/server_configuration_test.cc b/server/test/unit_tests/server_configuration_test.cc
similarity index 98%
rename from onnxruntime/test/server/unit_tests/server_configuration_test.cc
rename to server/test/unit_tests/server_configuration_test.cc
index 46e2b06b32fbe..578a550f4fa83 100644
--- a/onnxruntime/test/server/unit_tests/server_configuration_test.cc
+++ b/server/test/unit_tests/server_configuration_test.cc
@@ -4,7 +4,7 @@
 #include "gtest/gtest.h"
 
 #include "gmock/gmock.h"
-#include "server/server_configuration.h"
+#include "server_configuration.h"
 
 namespace onnxruntime {
 namespace server {
diff --git a/onnxruntime/test/server/unit_tests/test_main.cc b/server/test/unit_tests/test_main.cc
similarity index 81%
rename from onnxruntime/test/server/unit_tests/test_main.cc
rename to server/test/unit_tests/test_main.cc
index 09a5304d5f8d7..ee106cc3286b6 100644
--- a/onnxruntime/test/server/unit_tests/test_main.cc
+++ b/server/test/unit_tests/test_main.cc
@@ -2,15 +2,14 @@
 // Licensed under the MIT License.
 
 #include "gtest/gtest.h"
-#include "server/environment.h"
+#include "environment.h"
 #include "test_server_environment.h"
 
 GTEST_API_ int main(int argc, char** argv) {
   int status = 0;
-
   try {
+    testing::InitGoogleTest(&argc, argv);
     onnxruntime::server::test::TestServerEnvironment server_env{};
-    onnxruntime::test::TestEnvironment env{argc, argv, false};
     status = RUN_ALL_TESTS();
   } catch (const std::exception& ex) {
     std::cerr << ex.what();
diff --git a/onnxruntime/test/server/unit_tests/test_server_environment.cc b/server/test/unit_tests/test_server_environment.cc
similarity index 71%
rename from onnxruntime/test/server/unit_tests/test_server_environment.cc
rename to server/test/unit_tests/test_server_environment.cc
index bd42e2ba31b81..8ce32769d70a8 100644
--- a/onnxruntime/test/server/unit_tests/test_server_environment.cc
+++ b/server/test/unit_tests/test_server_environment.cc
@@ -1,7 +1,10 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
 #include "test_server_environment.h"
 #include <spdlog/spdlog.h>
 #include <spdlog/sinks/sink.h>
 #include <spdlog/sinks/stdout_sinks.h>
+#include <assert.h>
 
 namespace onnxruntime {
 namespace server {
@@ -9,9 +12,7 @@ namespace test {
 static std::unique_ptr<onnxruntime::server::ServerEnvironment> s_env;
 
 ServerEnvironment* ServerEnv() {
-  ORT_ENFORCE(s_env != nullptr,
-              "Need a TestServerEnvironment instance to provide the server environment.");
-
+  assert(s_env != nullptr);
   return s_env.get();
 }
 
@@ -19,7 +20,7 @@ TestServerEnvironment::TestServerEnvironment() {
   auto console = spdlog::stdout_logger_mt("console");
   spdlog::set_default_logger(console);
   spdlog::sink_ptr ptr = std::make_shared<spdlog::sinks::stdout_sink_st>();
-  s_env = onnxruntime::make_unique<onnxruntime::server::ServerEnvironment>(ORT_LOGGING_LEVEL_WARNING, spdlog::sinks_init_list{ptr});
+  s_env = std::make_unique<onnxruntime::server::ServerEnvironment>(ORT_LOGGING_LEVEL_WARNING, spdlog::sinks_init_list{ptr});
 }
 TestServerEnvironment::~TestServerEnvironment() {
   //destruct env to make sure the default logger is destoryed before the logger mutex.
@@ -28,4 +29,4 @@ TestServerEnvironment::~TestServerEnvironment() {
 
 }  // namespace test
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/test/server/unit_tests/test_server_environment.h b/server/test/unit_tests/test_server_environment.h
similarity index 71%
rename from onnxruntime/test/server/unit_tests/test_server_environment.h
rename to server/test/unit_tests/test_server_environment.h
index bfd8072f13d4b..e725d15d7f23c 100644
--- a/onnxruntime/test/server/unit_tests/test_server_environment.h
+++ b/server/test/unit_tests/test_server_environment.h
@@ -1,7 +1,8 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
 #pragma once
 
-#include "server/environment.h"
-#include "test/test_environment.h"
+#include "environment.h"
 
 namespace onnxruntime {
 namespace server {
@@ -17,4 +18,4 @@ class TestServerEnvironment {
 };
 }  // namespace test
 }  // namespace server
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/test/server/unit_tests/util_tests.cc b/server/test/unit_tests/util_tests.cc
similarity index 98%
rename from onnxruntime/test/server/unit_tests/util_tests.cc
rename to server/test/unit_tests/util_tests.cc
index c16f87c4a712b..48b17dd769d82 100644
--- a/onnxruntime/test/server/unit_tests/util_tests.cc
+++ b/server/test/unit_tests/util_tests.cc
@@ -3,8 +3,8 @@
 
 #include <google/protobuf/stubs/status.h>
 #include "gtest/gtest.h"
-#include "server/http/core/context.h"
-#include "server/http/util.h"
+#include "http/core/context.h"
+#include "http/util.h"
 
 namespace onnxruntime {
 namespace server {
diff --git a/onnxruntime/server/util.cc b/server/util.cc
similarity index 60%
rename from onnxruntime/server/util.cc
rename to server/util.cc
index a4dbc3ed291ae..2124e1f7bf8f5 100644
--- a/onnxruntime/server/util.cc
+++ b/server/util.cc
@@ -4,7 +4,6 @@
 #include <sstream>
 #include <google/protobuf/stubs/status.h>
 
-#include "core/common/status.h"
 #include "util.h"
 
 namespace onnxruntime {
@@ -15,23 +14,23 @@ namespace protobufutil = google::protobuf::util;
 protobufutil::Status GenerateProtobufStatus(const int& onnx_status, const std::string& message) {
   protobufutil::error::Code code = protobufutil::error::Code::UNKNOWN;
   switch (onnx_status) {
-    case onnxruntime::common::StatusCode::OK:
-    case onnxruntime::common::StatusCode::MODEL_LOADED:
+    case ORT_OK:
+    case ORT_MODEL_LOADED:
       code = protobufutil::error::Code::OK;
       break;
-    case onnxruntime::common::StatusCode::FAIL:
-    case onnxruntime::common::StatusCode::INVALID_ARGUMENT:
-    case onnxruntime::common::StatusCode::INVALID_PROTOBUF:
-    case onnxruntime::common::StatusCode::INVALID_GRAPH:
-    case onnxruntime::common::StatusCode::NO_SUCHFILE:
-    case onnxruntime::common::StatusCode::NO_MODEL:
+    case ORT_FAIL:
+    case ORT_INVALID_ARGUMENT:
+    case ORT_INVALID_PROTOBUF:
+    case ORT_INVALID_GRAPH:
+    case ORT_NO_SUCHFILE:
+    case ORT_NO_MODEL:
       code = protobufutil::error::Code::INVALID_ARGUMENT;
       break;
-    case onnxruntime::common::StatusCode::NOT_IMPLEMENTED:
+    case ORT_NOT_IMPLEMENTED:
       code = protobufutil::error::Code::UNIMPLEMENTED;
       break;
-    case onnxruntime::common::StatusCode::RUNTIME_EXCEPTION:
-    case onnxruntime::common::StatusCode::EP_FAIL:
+    case ORT_RUNTIME_EXCEPTION:
+    case ORT_EP_FAIL:
       code = protobufutil::error::Code::INTERNAL;
       break;
     default:
diff --git a/onnxruntime/server/util.h b/server/util.h
similarity index 79%
rename from onnxruntime/server/util.h
rename to server/util.h
index f136324e195ca..8ca204ad83bc8 100644
--- a/onnxruntime/server/util.h
+++ b/server/util.h
@@ -4,8 +4,9 @@
 #pragma once
 
 #include <google/protobuf/stubs/status.h>
+#include "onnxruntime_c_api.h"
+#include "onnxruntime_cxx_api.h"
 
-#include "core/common/status.h"
 
 namespace onnxruntime {
 namespace server {
@@ -39,8 +40,7 @@ class MemBufferArray {
 };
 
 google::protobuf::util::Status GenerateProtobufStatus(const int& onnx_status, const std::string& message);
-// Generate protobuf status from ONNX Runtime status
-google::protobuf::util::Status GenerateProtobufStatus(const onnxruntime::common::Status& onnx_status, const std::string& message);
+
 
 }  // namespace server
 }  // namespace onnxruntime
diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index b07d26408d15a..6b4a869c9f8d6 100755
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -95,12 +95,7 @@ def parse_arguments():
 
     # Build a shared lib
     parser.add_argument("--build_shared_lib", action='store_true', help="Build a shared library for the ONNXRuntime.")
-
-    # Build ONNX Runtime server
-    parser.add_argument("--build_server", action='store_true', help="Build server application for the ONNXRuntime.")
-    parser.add_argument("--enable_server_tests", action='store_true', help="Run server application tests.")
-    parser.add_argument("--enable_server_model_tests", action='store_true', help="Run server model tests.")
-
+        
     # Build options
     parser.add_argument("--cmake_extra_defines", nargs="+",
                         help="Extra definitions to pass to CMake during build system generation. " +
@@ -296,7 +291,7 @@ def generate_build_tree(cmake_path, source_dir, build_dir, cuda_home, cudnn_home
                  "-Donnxruntime_ENABLE_PYTHON=" + ("ON" if args.enable_pybind else "OFF"),
                  "-Donnxruntime_BUILD_CSHARP=" + ("ON" if args.build_csharp else "OFF"),
                  "-Donnxruntime_BUILD_JAVA=" + ("ON" if args.build_java else "OFF"),
-                 "-Donnxruntime_BUILD_SHARED_LIB=" + ("ON" if args.build_shared_lib or args.build_server else "OFF"),
+                 "-Donnxruntime_BUILD_SHARED_LIB=" + ("ON" if args.build_shared_lib else "OFF"),
                  "-Donnxruntime_USE_EIGEN_FOR_BLAS=" + ("OFF" if args.use_openblas else "ON"),
                  "-Donnxruntime_USE_OPENBLAS=" + ("ON" if args.use_openblas else "OFF"),
                  "-Donnxruntime_USE_DNNL=" + ("ON" if args.use_dnnl else "OFF"),
@@ -320,10 +315,9 @@ def generate_build_tree(cmake_path, source_dir, build_dir, cuda_home, cudnn_home
                  "-Donnxruntime_USE_TENSORRT=" + ("ON" if args.use_tensorrt else "OFF"),
                  "-Donnxruntime_TENSORRT_HOME=" + (tensorrt_home if args.use_tensorrt else ""),
                   # By default - we currently support only cross compiling for ARM/ARM64 (no native compilation supported through this script)
-                 "-Donnxruntime_CROSS_COMPILING=" + ("ON" if args.arm64 or args.arm else "OFF"),
-                 "-Donnxruntime_BUILD_SERVER=" + ("ON" if args.build_server else "OFF"),
+                 "-Donnxruntime_CROSS_COMPILING=" + ("ON" if args.arm64 or args.arm else "OFF"),                 
                   # nGraph and TensorRT providers currently only supports full_protobuf option.
-                 "-Donnxruntime_USE_FULL_PROTOBUF=" + ("ON" if args.use_full_protobuf or args.use_ngraph or args.use_tensorrt or args.build_server or args.gen_doc else "OFF"),
+                 "-Donnxruntime_USE_FULL_PROTOBUF=" + ("ON" if args.use_full_protobuf or args.use_ngraph or args.use_tensorrt or args.gen_doc else "OFF"),
                  "-Donnxruntime_DISABLE_CONTRIB_OPS=" + ("ON" if args.disable_contrib_ops else "OFF"),
                  "-Donnxruntime_MSVC_STATIC_RUNTIME=" + ("ON" if args.enable_msvc_static_runtime else "OFF"),
                  # enable pyop if it is nightly build
@@ -702,69 +696,6 @@ def nuphar_run_python_tests(build_dir, configs):
         run_subprocess([sys.executable, 'onnxruntime_test_python_nuphar.py'], cwd=cwd, dll_path=dll_path)
 
 
-def split_server_binary_and_symbol(build_dir, configs):
-    if is_windows():
-        # TODO: Windows support
-        pass
-    else:
-        for config in configs:
-            if config == 'RelWithDebInfo':
-                config_build_dir = get_config_build_dir(build_dir, config)
-                run_subprocess(['objcopy', '--only-keep-debug', 'onnxruntime_server', 'onnxruntime_server.symbol'], cwd=config_build_dir)
-                run_subprocess(['strip', '--strip-debug', '--strip-unneeded', 'onnxruntime_server'], cwd=config_build_dir)
-                run_subprocess(['objcopy', '--add-gnu-debuglink=onnxruntime_server.symbol', 'onnxruntime_server'], cwd=config_build_dir)
-                libonnx = glob.glob(os.path.join(config_build_dir, "libonnxruntime.so.*"))
-                if len(libonnx) != 1 :
-                    raise ValueError("Too many libonxruntime.so.*")
-                libonnx = libonnx[0]
-                run_subprocess(['objcopy', '--only-keep-debug', libonnx, libonnx+'.symbol'], cwd=config_build_dir)
-                run_subprocess(['strip', '--strip-debug', libonnx], cwd=config_build_dir)
-                run_subprocess(['objcopy', '--add-gnu-debuglink={}.symbol'.format(libonnx), libonnx], cwd=config_build_dir)
-
-
-def run_server_tests(build_dir, configs):
-    pip_freeze_result = run_subprocess([sys.executable, '-m', 'pip', 'freeze'], capture=True).stdout
-    installed_packages = [r.decode().split('==')[0] for r in pip_freeze_result.split()]
-    if not (('requests' in installed_packages) and ('protobuf' in installed_packages) and ('numpy' in installed_packages) and ('grpcio' in installed_packages)):
-        if hasattr(sys, 'real_prefix'):
-            # In virtualenv
-            run_subprocess([sys.executable, '-m', 'pip', 'install', '--trusted-host', 'files.pythonhosted.org', 'requests', 'protobuf', 'numpy', 'grpcio'])
-        else:
-            # Outside virtualenv
-            run_subprocess([sys.executable, '-m', 'pip', 'install', '--user', '--trusted-host', 'files.pythonhosted.org', 'requests', 'protobuf', 'numpy', 'grpcio'])
-    for config in configs:
-        config_build_dir = get_config_build_dir(build_dir, config)
-        if is_windows():
-            server_app_path = os.path.join(config_build_dir, config, 'onnxruntime_server.exe')
-            python_package_path = os.path.join(config_build_dir, config)
-        else:
-            server_app_path = os.path.join(config_build_dir, 'onnxruntime_server')
-            python_package_path = config_build_dir
-        server_test_folder = os.path.join(config_build_dir, 'server_test')
-        server_test_model_folder = os.path.join(build_dir, 'models', 'opset8', 'test_mnist')
-        server_test_data_folder = os.path.join(config_build_dir, 'testdata', 'server')
-        run_subprocess([sys.executable, 'test_main.py', server_app_path, server_test_model_folder, server_test_data_folder, python_package_path, server_test_folder], cwd=server_test_folder, dll_path=None)
-
-
-def run_server_model_tests(build_dir, configs):
-    for config in configs:
-        config_build_dir = get_config_build_dir(build_dir, config)
-        server_test_folder = os.path.join(config_build_dir, 'server_test')
-        server_test_data_folder = os.path.join(config_build_dir, 'server_test_data')
-
-        if is_windows():
-            server_app_path = os.path.join(config_build_dir, config, 'onnxruntime_server.exe')
-            test_raw_data_folder = os.path.join(config_build_dir, 'models')
-            python_package_path = os.path.join(config_build_dir, config)
-        else:
-            server_app_path = os.path.join(config_build_dir, 'onnxruntime_server')
-            test_raw_data_folder = os.path.join(build_dir, 'models')
-            python_package_path = config_build_dir
-
-        run_subprocess([sys.executable, 'model_zoo_data_prep.py', test_raw_data_folder, server_test_data_folder, python_package_path, server_test_folder], cwd=server_test_folder, dll_path=None)
-        run_subprocess([sys.executable, 'model_zoo_tests.py', server_app_path, test_raw_data_folder, server_test_data_folder, python_package_path, server_test_folder], cwd=server_test_folder, dll_path=None)
-
-
 def build_python_wheel(source_dir, build_dir, configs, use_cuda, use_ngraph, use_tensorrt, use_openvino, use_nuphar, nightly_build = False):
     for config in configs:
         cwd = get_config_build_dir(build_dir, config)
@@ -884,7 +815,7 @@ def main():
     if args.use_tensorrt:
         args.use_cuda = True
 
-    if args.build_wheel or args.enable_server_model_tests:
+    if args.build_wheel:
         args.enable_pybind = True
 
     if args.build_csharp or args.build_java:
@@ -1026,13 +957,6 @@ def main():
         if args.enable_pybind and not args.skip_onnx_tests and args.use_nuphar:
             nuphar_run_python_tests(build_dir, configs)
 
-    if args.build_server:
-        split_server_binary_and_symbol(build_dir, configs)
-        if args.enable_server_tests:
-            run_server_tests(build_dir, configs)
-        if args.enable_server_model_tests:
-            run_server_model_tests(build_dir, configs)
-
     if args.build:
         if args.build_wheel:
             nightly_build = bool(os.getenv('NIGHTLY_BUILD') == '1')
diff --git a/tools/ci_build/github/azure-pipelines/linux-ort-srv-ci-pipeline.yml b/tools/ci_build/github/azure-pipelines/linux-ort-srv-ci-pipeline.yml
index 2786df3c4e905..6d9b79a274890 100644
--- a/tools/ci_build/github/azure-pipelines/linux-ort-srv-ci-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/linux-ort-srv-ci-pipeline.yml
@@ -14,6 +14,12 @@ jobs:
       continueOnError: true
       condition: always()
 
+    - task: CmdLine@2
+      displayName: 'Build docker image'
+      inputs:
+        script: docker build --pull -t onnxruntime-server-ubuntu16.04 --build-arg BUILD_USER=onnxruntimedev --build-arg BUILD_UID=$(id -u) --build-arg OS_VERSION=16.04 --build-arg PYTHON_VERSION=3.5 -f Dockerfile.ubuntu_server .         
+        workingDirectory: $(Build.SourcesDirectory)/tools/ci_build/github/linux/docker
+
     - task: CmdLine@2
       displayName: 'Download azcopy'
       inputs:
@@ -30,8 +36,11 @@ jobs:
         pythonInterpreter: '/usr/bin/python3'
         workingDirectory: $(Build.BinariesDirectory)
 
-    - script: 'tools/ci_build/github/linux/server_run_dockerbuild.sh -o ubuntu16.04 -d cpu -r $(Build.BinariesDirectory) -x "--config Debug Release --build_server --use_openmp --use_full_protobuf --enable_server_tests"'
-      displayName: 'Run build script'
+    - task: CmdLine@2
+      displayName: 'Run docker image'
+      inputs:
+        script: docker run --rm --volume $(Build.SourcesDirectory)/server:/onnxruntime_src --volume $(Build.BinariesDirectory):/build onnxruntime-server-ubuntu16.04 /bin/bash /onnxruntime_src/ci/run.sh        
+        workingDirectory: $(Build.SourcesDirectory)/tools/ci_build/github/linux/docker
 
     - task: ms.vss-governance-buildtask.governance-build-task-component-detection.ComponentGovernanceComponentDetection@0
       displayName: 'Component Detection'
diff --git a/tools/ci_build/github/linux/docker/Dockerfile.ubuntu_server b/tools/ci_build/github/linux/docker/Dockerfile.ubuntu_server
index ce75f212d370f..cc65a4b294ca4 100644
--- a/tools/ci_build/github/linux/docker/Dockerfile.ubuntu_server
+++ b/tools/ci_build/github/linux/docker/Dockerfile.ubuntu_server
@@ -4,7 +4,7 @@ FROM ubuntu:${OS_VERSION}
 ARG PYTHON_VERSION=3.5
 
 ADD scripts /tmp/scripts
-RUN /tmp/scripts/install_ubuntu.sh -p $PYTHON_VERSION && /tmp/scripts/install_deps.sh -p $PYTHON_VERSION && /tmp/scripts/install_server_deps.sh && rm -rf /tmp/scripts
+RUN /tmp/scripts/install_ubuntu.sh -p $PYTHON_VERSION && /tmp/scripts/install_server_deps.sh && rm -rf /tmp/scripts
 
 WORKDIR /root
 
diff --git a/tools/ci_build/github/linux/docker/scripts/install_server_deps.sh b/tools/ci_build/github/linux/docker/scripts/install_server_deps.sh
index d471064191684..7d30f542625f3 100755
--- a/tools/ci_build/github/linux/docker/scripts/install_server_deps.sh
+++ b/tools/ci_build/github/linux/docker/scripts/install_server_deps.sh
@@ -3,12 +3,77 @@ set -e
 
 SYS_LONG_BIT=$(getconf LONG_BIT)
 
-echo "Installing Go"
+apt-get remove -y libprotobuf-dev libprotobuf9v5 protobuf-compiler libprotobuf-lite9v5
+apt-get update && apt-get install -y --no-install-recommends libre2-dev
+rm -rf /var/lib/apt/lists/*
+
 if [ $SYS_LONG_BIT = "64" ]; then
+  echo "Installing Go"
   mkdir -p /tmp/go
   cd /tmp/go
   wget https://dl.google.com/go/go1.12.6.linux-amd64.tar.gz
   tar -C /usr/local -vzxf /tmp/go/go1.12.6.linux-amd64.tar.gz
+
+  echo "Installing CMAKE"
+  aria2c https://github.com/Kitware/CMake/releases/download/v3.16.2/cmake-3.16.2-Linux-x86_64.tar.gz
+  tar -zxf cmake-3.16.2-Linux-x86_64.tar.gz --strip=1 -C /usr
+
+  echo "Installing onnxruntime"
+  aria2c https://github.com/microsoft/onnxruntime/releases/download/v1.1.0/onnxruntime-linux-x64-1.1.0.tgz
+  tar -zxf onnxruntime-linux-x64-1.1.0.tgz --strip=1
+  cp -r include/* /usr/include 
+  cp -r lib/* /usr/lib
+  ldconfig /usr/lib
 fi
 
+echo "Installing googletest from source"
+cd /tmp
+git clone https://github.com/google/googletest.git
+cd googletest
+git checkout release-1.10.0
+mkdir b
+cd b
+cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr/local -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+make -j$(getconf _NPROCESSORS_ONLN)
+make install
+cd /tmp
+rm -rf /tmp/googletest
+
+
+echo "Installing protobuf from source"
+git clone https://github.com/protocolbuffers/protobuf.git
+cd protobuf
+git checkout v3.11.2
+mkdir b
+cd b
+cmake ../cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_POSITION_INDEPENDENT_CODE=ON -Dprotobuf_BUILD_TESTS=OFF -DCMAKE_BUILD_TYPE=Debug
+make -j$(getconf _NPROCESSORS_ONLN)
+make install
+cd ..
+mkdir b2
+cd b2
+cmake ../cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_POSITION_INDEPENDENT_CODE=ON -Dprotobuf_BUILD_TESTS=OFF -DCMAKE_BUILD_TYPE=Release
+make -j$(getconf _NPROCESSORS_ONLN)
+make install
+cd /tmp
+rm -rf /tmp/protobuf
 
+echo "Installing grpc"
+git clone -b v1.26.0 https://github.com/grpc/grpc
+cd grpc
+git submodule update --init
+mkdir -p "third_party/cares/cares/cmake/build"
+pushd "third_party/cares/cares/cmake/build"
+cmake -DCMAKE_BUILD_TYPE=Release ../..
+make -j$(getconf _NPROCESSORS_ONLN) install
+popd
+mkdir -p cmake/build
+cd cmake/build
+cmake ../.. -DCMAKE_BUILD_TYPE=Release -DgRPC_CARES_PROVIDER=package    \
+           -DgRPC_PROTOBUF_PROVIDER=package \
+           -DgRPC_SSL_PROVIDER=package      \
+           -DgRPC_ZLIB_PROVIDER=package
+make -j$(getconf _NPROCESSORS_ONLN)
+make install
+cd /tmp
+rm -rf /tmp/grpc
diff --git a/tools/ci_build/github/linux/server_run_build.sh b/tools/ci_build/github/linux/server_run_build.sh
deleted file mode 100755
index 9ca7ccd3b9ec2..0000000000000
--- a/tools/ci_build/github/linux/server_run_build.sh
+++ /dev/null
@@ -1,29 +0,0 @@
-#!/bin/bash
-set -e -o -x
-
- id
-
- SCRIPT_DIR="$( dirname "${BASH_SOURCE[0]}" )"
-
- while getopts c:d:x: parameter_Option
-do case "${parameter_Option}"
-in
-d) BUILD_DEVICE=${OPTARG};;
-x) BUILD_EXTR_PAR=${OPTARG};;
-esac
-done
-
- if [ $BUILD_DEVICE = "gpu" ]; then
-    _CUDNN_VERSION=$(echo $CUDNN_VERSION | cut -d. -f1-2)
-    python3 $SCRIPT_DIR/../../build.py --build_dir /build \
-        --config Debug Release \
-        --skip_submodule_sync --enable_onnx_tests \
-        --parallel --build_shared_lib \
-        --use_cuda --use_openmp \
-        --cuda_home /usr/local/cuda \
-        --cudnn_home /usr/local/cudnn-$_CUDNN_VERSION/cuda --build_shared_lib $BUILD_EXTR_PAR
-else
-    python3 $SCRIPT_DIR/../../build.py --build_dir /build \
-        --skip_submodule_sync  \
-        --parallel $BUILD_EXTR_PAR
-fi
\ No newline at end of file
diff --git a/tools/ci_build/github/linux/server_run_dockerbuild.sh b/tools/ci_build/github/linux/server_run_dockerbuild.sh
deleted file mode 100755
index 41018d2362317..0000000000000
--- a/tools/ci_build/github/linux/server_run_dockerbuild.sh
+++ /dev/null
@@ -1,81 +0,0 @@
-#!/bin/bash
-set -e -o -x
-
-SCRIPT_DIR="$( dirname "${BASH_SOURCE[0]}" )"
-SOURCE_ROOT=$(realpath $SCRIPT_DIR/../../../../)
-CUDA_VER=cuda10.0-cudnn7.3
-
-while getopts c:o:d:r:p:x:a: parameter_Option
-do case "${parameter_Option}"
-in
-#android, ubuntu16.04
-o) BUILD_OS=${OPTARG};;
-#cpu, gpu, tensorrt
-d) BUILD_DEVICE=${OPTARG};;
-r) BUILD_DIR=${OPTARG};;
-#python version: 3.6 3.7 (absence means default 3.5)
-p) PYTHON_VER=${OPTARG};;
-# "--build_wheel --use_openblas"
-x) BUILD_EXTR_PAR=${OPTARG};;
-# "cuda10.0-cudnn7.3, cuda9.1-cudnn7.1"
-c) CUDA_VER=${OPTARG};;
-# x86 or other, only for ubuntu16.04 os
-a) BUILD_ARCH=${OPTARG};;
-esac
-done
-
-EXIT_CODE=1
-PYTHON_VER=${PYTHON_VER:=3.5}
-echo "bo=$BUILD_OS bd=$BUILD_DEVICE bdir=$BUILD_DIR pv=$PYTHON_VER bex=$BUILD_EXTR_PAR"
-
-IMAGE=ubuntu16.04
-
-cd $SCRIPT_DIR/docker
-docker build --pull -t "onnxruntime-server-$IMAGE" --build-arg BUILD_USER=onnxruntimedev --build-arg BUILD_UID=$(id -u) --build-arg OS_VERSION=16.04 --build-arg PYTHON_VERSION=${PYTHON_VER} -f Dockerfile.ubuntu_server .
-
-
-set +e
-mkdir -p ~/.cache/onnxruntime
-mkdir -p ~/.onnx
-mkdir -p ~/.cache/go
-
-if [ -z "$NIGHTLY_BUILD" ]; then
-set NIGHTLY_BUILD=0
-fi
-
-if [ $BUILD_DEVICE = "cpu" ] || [ $BUILD_DEVICE = "ngraph" ]; then
-    docker rm -f "onnxruntime-$BUILD_DEVICE" || true
-    docker run -h $HOSTNAME \
-        --name "onnxruntime-$BUILD_DEVICE" \
-        --volume "$SOURCE_ROOT:/onnxruntime_src" \
-        --volume "$BUILD_DIR:/build" \
-        --volume "$HOME/.cache/onnxruntime:/home/onnxruntimedev/.cache/onnxruntime" \
-        --volume "$HOME/.onnx:/home/onnxruntimedev/.onnx" \
-        --volume "$HOME/.cache/go:/home/onnxruntimedev/.cache/go" \
-        -e NIGHTLY_BUILD \
-        -e GOCACHE=/home/onnxruntimedev/.cache/go \
-        "onnxruntime-server-$IMAGE" \
-        /bin/bash /onnxruntime_src/tools/ci_build/github/linux/server_run_build.sh \
-         -d $BUILD_DEVICE -x "$BUILD_EXTR_PAR" -o $BUILD_OS &
-else
-    docker rm -f "onnxruntime-$BUILD_DEVICE" || true
-    nvidia-docker run --rm -h $HOSTNAME \
-        --rm \
-        --name "onnxruntime-$BUILD_DEVICE" \
-        --volume "$SOURCE_ROOT:/onnxruntime_src" \
-        --volume "$BUILD_DIR:/build" \
-        --volume "$HOME/.cache/onnxruntime:/home/onnxruntimedev/.cache/onnxruntime" \
-        --volume "$HOME/.onnx:/home/onnxruntimedev/.onnx" \
-        --volume "$HOME/.cache/go:/home/onnxruntimedev/.cache/go" \
-        -e NIGHTLY_BUILD \
-        -e GOCACHE=/home/onnxruntimedev/.cache/go \
-        "onnxruntime-server-$IMAGE" \
-        /bin/bash /onnxruntime_src/tools/ci_build/github/linux/server_run_build.sh \
-        -d $BUILD_DEVICE -x "$BUILD_EXTR_PAR" -o $BUILD_OS &
-fi
-wait $!
-
-EXIT_CODE=$?
-
-set -e
-exit $EXIT_CODE

From b40a85a0e8f89928b965a5ed5e90e158cd642355 Mon Sep 17 00:00:00 2001
From: Changyoung Koh <gkcy1019@gmail.com>
Date: Sat, 28 Dec 2019 05:17:06 +0900
Subject: [PATCH 14/75] Fix build for Python 3.8 (#2747)

* Fix build for Python 3.8
---
 tools/ci_build/build.py | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index 6b4a869c9f8d6..f420cb13cda81 100755
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -167,8 +167,17 @@ def resolve_executable_path(command_or_path):
 def is_windows():
     return sys.platform.startswith("win")
 
+def get_linux_distro():
+    try:
+        with open('/etc/os-release', 'r') as f:
+            dist_info = dict(line.strip().split('=', 1) for line in f.readlines())
+        return dist_info.get('NAME', '').strip('"'), dist_info.get('VERSION', '').strip('"')
+    except:
+        return '', ''
+
 def is_ubuntu_1604():
-    return platform.linux_distribution()[0] == 'Ubuntu' and platform.linux_distribution()[1] == '16.04'
+    dist, ver = get_linux_distro()
+    return dist == 'Ubuntu' and ver.startswith('16.04')
 
 def get_config_build_dir(build_dir, config):
     # build directory per configuration

From 90b708f8a9b8191593ee0d9dbe4e5f8037921749 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 27 Dec 2019 18:28:18 -0800
Subject: [PATCH 15/75] Update protobuf to 3.11.2 (#1928)

Update protobuf to 3.11.2 (#1928)
---
 cgmanifest.json                               |   2 +-
 cmake/CMakeLists.txt                          |  16 +-
 cmake/external/protobuf                       |   2 +-
 cmake/onnx/CMakeLists.txt                     |   6 +-
 cmake/onnxruntime_unittests.cmake             |  17 +-
 cmake/patches/ngraph/ngraph_protobuf.patch    |   2 +-
 .../Microsoft.ML.OnnxRuntime.csproj           |   2 +-
 ...rosoft.ML.OnnxRuntime.EndToEndTests.csproj |   2 +-
 .../Microsoft.ML.OnnxRuntime.Tests.csproj     |   2 +-
 .../Microsoft.ML.OnnxRuntime.Tests/OnnxMl.cs  | 144 +++---
 onnxruntime/core/framework/path_lib.h         |  10 +-
 onnxruntime/core/graph/model.cc               |   2 +-
 onnxruntime/core/session/onnxruntime_c_api.cc |  17 +-
 .../core/util/protobuf_parsing_utils.cc       | 481 ------------------
 .../core/util/protobuf_parsing_utils.h        |  62 ---
 onnxruntime/test/ir/onnx_model_test.cc        |   4 +-
 onnxruntime/test/onnx/TestCase.cc             |   8 +-
 onnxruntime/test/onnx/main.cc                 |  34 +-
 onnxruntime/test/onnx/runner.cc               |   4 +
 onnxruntime/test/onnx/runner.h                |   1 +
 onnxruntime/test/util/compare_ortvalue.cc     |  28 +
 21 files changed, 182 insertions(+), 664 deletions(-)
 delete mode 100644 onnxruntime/core/util/protobuf_parsing_utils.cc

diff --git a/cgmanifest.json b/cgmanifest.json
index cdcfeefaa2ca1..782db27f7c8d0 100644
--- a/cgmanifest.json
+++ b/cgmanifest.json
@@ -247,7 +247,7 @@
          "component": {
             "type": "git",
             "git": {
-               "commitHash": "48cb18e5c419ddd23d9badcfe4e9df7bde1979b2",
+               "commitHash": "fe1790ca0df67173702f70d5646b82f48f412b99",
                "repositoryUrl": "https://github.com/protocolbuffers/protobuf.git"
             }
          }
diff --git a/cmake/CMakeLists.txt b/cmake/CMakeLists.txt
index 655b14792f731..7da78a539ddd8 100644
--- a/cmake/CMakeLists.txt
+++ b/cmake/CMakeLists.txt
@@ -270,10 +270,14 @@ list(APPEND CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/external)
 
 
 # use protobuf as a submodule
+if (CMAKE_SYSTEM_NAME STREQUAL "Android")
+  set(protobuf_BUILD_PROTOC_BINARIES OFF CACHE BOOL "Build protobuf tests" FORCE)
+endif()
+
 add_subdirectory(${PROJECT_SOURCE_DIR}/external/protobuf/cmake EXCLUDE_FROM_ALL)
 set_target_properties(libprotobuf PROPERTIES FOLDER "External/Protobuf")
 set_target_properties(libprotobuf-lite PROPERTIES FOLDER "External/Protobuf")
-set_target_properties(libprotoc PROPERTIES FOLDER "External/Protobuf")
+set_target_properties(libprotoc PROPERTIES FOLDER "External/Protobuf")	
 set_target_properties(protoc PROPERTIES FOLDER "External/Protobuf")
 if (onnxruntime_USE_FULL_PROTOBUF)
   add_library(protobuf::libprotobuf ALIAS libprotobuf)
@@ -286,7 +290,6 @@ if(UNIX AND onnxruntime_ENABLE_LTO)
   #https://github.com/protocolbuffers/protobuf/issues/5923
   target_link_options(protoc PRIVATE "-Wl,--no-as-needed")
 endif()
-
 include(protobuf_function.cmake)
 
 if (onnxruntime_DISABLE_CONTRIB_OPS)
@@ -413,6 +416,14 @@ if (onnxruntime_USE_TVM)
   list(APPEND onnxruntime_EXTERNAL_DEPENDENCIES tvm nnvm_compiler)
 endif()
 
+if (APPLE)
+  #onnx/onnx/proto_utils.h:34:16: error: 'SetTotalBytesLimit' is deprecated: Please use the single 
+  #parameter version of SetTotalBytesLimit(). The second parameter is ignored. 
+  #  coded_stream.SetTotalBytesLimit((2048LL << 20) - 1, 512LL << 20);
+  #TODO: fix the warning in ONNX and re-enable this flag
+  string(APPEND CMAKE_CXX_FLAGS " -Wno-deprecated")
+  string(APPEND CMAKE_C_FLAGS " -Wno-deprecated")
+endif()
 # ONNX
 add_subdirectory(onnx)
 
@@ -487,6 +498,7 @@ else()
   endif()
   check_cxx_compiler_flag(-Wunused-but-set-variable HAS_UNUSED_BUT_SET_VARIABLE)
   check_cxx_compiler_flag(-Wunused-parameter HAS_UNUSED_PARAMETER)
+  check_cxx_compiler_flag(-Wunused-variable HAS_UNUSED_VARIABLE)
   check_cxx_compiler_flag(-Wcast-function-type HAS_CAST_FUNCTION_TYPE)
   check_cxx_compiler_flag(-Wparentheses HAS_PARENTHESES)
   check_cxx_compiler_flag(-Wuseless-cast HAS_USELESS_CAST)
diff --git a/cmake/external/protobuf b/cmake/external/protobuf
index 48cb18e5c419d..7bb8b108d1625 160000
--- a/cmake/external/protobuf
+++ b/cmake/external/protobuf
@@ -1 +1 @@
-Subproject commit 48cb18e5c419ddd23d9badcfe4e9df7bde1979b2
+Subproject commit 7bb8b108d16252d0ed053673d70ea6d2020ec7ff
diff --git a/cmake/onnx/CMakeLists.txt b/cmake/onnx/CMakeLists.txt
index 3c7e6b3d7064e..1048c90bd2c5c 100644
--- a/cmake/onnx/CMakeLists.txt
+++ b/cmake/onnx/CMakeLists.txt
@@ -6,7 +6,11 @@ target_include_directories(onnx_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprot
 target_compile_definitions(onnx_proto PUBLIC $<TARGET_PROPERTY:protobuf::libprotobuf,INTERFACE_COMPILE_DEFINITIONS>)
 onnxruntime_protobuf_generate(APPEND_PATH IMPORT_DIRS ${ONNXRUNTIME_ROOT}/core/protobuf TARGET onnx_proto)
 if (WIN32)
-  target_compile_options(onnx_proto PRIVATE /wd4146) # unary minus operator applied to unsigned type
+  target_compile_options(onnx_proto PRIVATE "/wd4146" "/wd4125" "/wd4456" "/wd4267")
+else()
+  if(HAS_UNUSED_VARIABLE)
+    target_compile_options(onnx_proto PRIVATE "-Wno-unused-variable")
+  endif()
 endif()
 # Cpp Tests were added and they require googletest
 # since we have our own copy, try using that
diff --git a/cmake/onnxruntime_unittests.cmake b/cmake/onnxruntime_unittests.cmake
index d3564feffb719..6bb562abf8be3 100644
--- a/cmake/onnxruntime_unittests.cmake
+++ b/cmake/onnxruntime_unittests.cmake
@@ -480,16 +480,23 @@ if(WIN32)
 endif()
 
 add_library(onnx_test_data_proto ${TEST_SRC_DIR}/proto/tml.proto)
-if(WIN32)
-    target_compile_options(onnx_test_data_proto PRIVATE "/wd4125" "/wd4456" "/wd4100")
-endif()
 add_dependencies(onnx_test_data_proto onnx_proto ${onnxruntime_EXTERNAL_DEPENDENCIES})
 
-if(NOT WIN32)
+if(WIN32)
+  target_compile_options(onnx_test_data_proto PRIVATE "/wd4125" "/wd4456" "/wd4100" "/wd4267")  
+else()
   if(HAS_UNUSED_PARAMETER)
-    set_source_files_properties(${CMAKE_CURRENT_BINARY_DIR}/tml.pb.cc PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
+    target_compile_options(onnx_test_data_proto PRIVATE "-Wno-unused-parameter")
+  endif()
+  if(HAS_UNUSED_VARIABLE)
+    target_compile_options(onnx_test_data_proto PRIVATE "-Wno-unused-variable")
+  endif()
+  if(HAS_UNUSED_BUT_SET_VARIABLE)    
+    target_compile_options(onnx_test_data_proto PRIVATE "-Wno-unused-but-set-variable")
   endif()
 endif()
+add_dependencies(onnx_test_data_proto onnx_proto ${onnxruntime_EXTERNAL_DEPENDENCIES})
+
 onnxruntime_add_include_to_target(onnx_test_data_proto onnx_proto)
 target_include_directories(onnx_test_data_proto PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_BINARY_DIR}/onnx)
 set_target_properties(onnx_test_data_proto PROPERTIES FOLDER "ONNXRuntimeTest")
diff --git a/cmake/patches/ngraph/ngraph_protobuf.patch b/cmake/patches/ngraph/ngraph_protobuf.patch
index 87b7b5bd39f19..57bf85a881d7a 100644
--- a/cmake/patches/ngraph/ngraph_protobuf.patch
+++ b/cmake/patches/ngraph/ngraph_protobuf.patch
@@ -11,7 +11,7 @@ index 32217f5d6..d33e9c631 100644
  # This version of PROTOBUF is required by Microsoft ONNX Runtime.
  set(NGRAPH_PROTOBUF_GIT_REPO_URL "https://github.com/protocolbuffers/protobuf")
 -set(NGRAPH_PROTOBUF_GIT_TAG "v3.5.2")
-+set(NGRAPH_PROTOBUF_GIT_TAG "v3.6.1")
++set(NGRAPH_PROTOBUF_GIT_TAG "v3.11.2")
  
  if (WIN32)
      ExternalProject_Add(
diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj b/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj
index ea268580165f4..c4cc6a526da36 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj
@@ -173,7 +173,7 @@
   </Target>
 
   <ItemGroup>
-    <PackageReference Include="Microsoft.SourceLink.GitHub" Version="1.0.0-beta-63127-02" PrivateAssets="All"/>
+    <PackageReference Include="Microsoft.SourceLink.GitHub" Version="1.0.0" PrivateAssets="All"/>
     <PackageReference Include="System.Memory" Version="4.5.3" />
   </ItemGroup>
 
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests/Microsoft.ML.OnnxRuntime.EndToEndTests.csproj b/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests/Microsoft.ML.OnnxRuntime.EndToEndTests.csproj
index e2ffae1c4cb4e..19ba6724e8ca8 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests/Microsoft.ML.OnnxRuntime.EndToEndTests.csproj
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.EndToEndTests/Microsoft.ML.OnnxRuntime.EndToEndTests.csproj
@@ -13,7 +13,7 @@
     <PackageName Condition="'$(PackageName)' == ''">Microsoft.ML.OnnxRuntime</PackageName>
   </PropertyGroup>
   <ItemGroup>
-    <PackageReference Include="Google.Protobuf" Version="3.6.1" />
+    <PackageReference Include="Google.Protobuf" Version="3.11.2" />
     <PackageReference Include="Microsoft.NET.Test.Sdk" Version="15.8.0" />
     <PackageReference Include="xunit" Version="2.4.0" />
     <PackageReference Include="xunit.runner.visualstudio" Version="2.4.0" />
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/Microsoft.ML.OnnxRuntime.Tests.csproj b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/Microsoft.ML.OnnxRuntime.Tests.csproj
index 5a74dd3674fb0..04c8a1cffa639 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/Microsoft.ML.OnnxRuntime.Tests.csproj
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/Microsoft.ML.OnnxRuntime.Tests.csproj
@@ -47,7 +47,7 @@
   <ItemGroup>
     <PackageReference Include="Microsoft.NETCore.Targets" Version="2.1.0"/>
     <PackageReference Include="System.Runtime.CompilerServices.Unsafe" Version="4.5.2"/>
-    <PackageReference Include="Google.Protobuf" Version="3.6.1" />
+    <PackageReference Include="Google.Protobuf" Version="3.11.2" />
     <PackageReference Include="Microsoft.NET.Test.Sdk" Version="16.2.0" />
     <PackageReference Include="xunit" Version="2.4.1" />
     <PackageReference Include="xunit.runner.visualstudio" Version="2.4.1" />
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/OnnxMl.cs b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/OnnxMl.cs
index dc04a19a2de5a..57df68e12e6d5 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/OnnxMl.cs
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/OnnxMl.cs
@@ -109,24 +109,24 @@ static OnnxMlReflection() {
             "EhAKDEVYUEVSSU1FTlRBTBAAEgoKBlNUQUJMRRABYgZwcm90bzM="));
       descriptor = pbr::FileDescriptor.FromGeneratedCode(descriptorData,
           new pbr::FileDescriptor[] { },
-          new pbr::GeneratedClrTypeInfo(new[] {typeof(global::Onnx.Version), typeof(global::Onnx.OperatorStatus), }, new pbr::GeneratedClrTypeInfo[] {
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.AttributeProto), global::Onnx.AttributeProto.Parser, new[]{ "Name", "RefAttrName", "DocString", "Type", "F", "I", "S", "T", "G", "SparseTensor", "Floats", "Ints", "Strings", "Tensors", "Graphs", "SparseTensors" }, null, new[]{ typeof(global::Onnx.AttributeProto.Types.AttributeType) }, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.ValueInfoProto), global::Onnx.ValueInfoProto.Parser, new[]{ "Name", "Type", "DocString" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.NodeProto), global::Onnx.NodeProto.Parser, new[]{ "Input", "Output", "Name", "OpType", "Domain", "Attribute", "DocString" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.ModelProto), global::Onnx.ModelProto.Parser, new[]{ "IrVersion", "OpsetImport", "ProducerName", "ProducerVersion", "Domain", "ModelVersion", "DocString", "Graph", "Functions", "MetadataProps" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.StringStringEntryProto), global::Onnx.StringStringEntryProto.Parser, new[]{ "Key", "Value" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorAnnotation), global::Onnx.TensorAnnotation.Parser, new[]{ "TensorName", "QuantParameterTensorNames" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.GraphProto), global::Onnx.GraphProto.Parser, new[]{ "Node", "Name", "Initializer", "SparseInitializer", "DocString", "Input", "Output", "ValueInfo", "QuantizationAnnotation" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorProto), global::Onnx.TensorProto.Parser, new[]{ "Dims", "DataType", "Segment", "FloatData", "Int32Data", "StringData", "Int64Data", "Name", "DocString", "RawData", "ExternalData", "DataLocation", "DoubleData", "Uint64Data" }, null, new[]{ typeof(global::Onnx.TensorProto.Types.DataType), typeof(global::Onnx.TensorProto.Types.DataLocation) }, new pbr::GeneratedClrTypeInfo[] { new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorProto.Types.Segment), global::Onnx.TensorProto.Types.Segment.Parser, new[]{ "Begin", "End" }, null, null, null)}),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.SparseTensorProto), global::Onnx.SparseTensorProto.Parser, new[]{ "Values", "Indices", "Dims" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorShapeProto), global::Onnx.TensorShapeProto.Parser, new[]{ "Dim" }, null, null, new pbr::GeneratedClrTypeInfo[] { new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorShapeProto.Types.Dimension), global::Onnx.TensorShapeProto.Types.Dimension.Parser, new[]{ "DimValue", "DimParam", "Denotation" }, new[]{ "Value" }, null, null)}),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto), global::Onnx.TypeProto.Parser, new[]{ "TensorType", "SequenceType", "MapType", "SparseTensorType", "OpaqueType", "Denotation" }, new[]{ "Value" }, null, new pbr::GeneratedClrTypeInfo[] { new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Tensor), global::Onnx.TypeProto.Types.Tensor.Parser, new[]{ "ElemType", "Shape" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Sequence), global::Onnx.TypeProto.Types.Sequence.Parser, new[]{ "ElemType" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Map), global::Onnx.TypeProto.Types.Map.Parser, new[]{ "KeyType", "ValueType" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.SparseTensor), global::Onnx.TypeProto.Types.SparseTensor.Parser, new[]{ "ElemType", "Shape" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Opaque), global::Onnx.TypeProto.Types.Opaque.Parser, new[]{ "Domain", "Name" }, null, null, null)}),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.OperatorSetIdProto), global::Onnx.OperatorSetIdProto.Parser, new[]{ "Domain", "Version" }, null, null, null),
-            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.FunctionProto), global::Onnx.FunctionProto.Parser, new[]{ "Name", "SinceVersion", "Status", "Input", "Output", "Attribute", "Node", "DocString" }, null, null, null)
+          new pbr::GeneratedClrTypeInfo(new[] {typeof(global::Onnx.Version), typeof(global::Onnx.OperatorStatus), }, null, new pbr::GeneratedClrTypeInfo[] {
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.AttributeProto), global::Onnx.AttributeProto.Parser, new[]{ "Name", "RefAttrName", "DocString", "Type", "F", "I", "S", "T", "G", "SparseTensor", "Floats", "Ints", "Strings", "Tensors", "Graphs", "SparseTensors" }, null, new[]{ typeof(global::Onnx.AttributeProto.Types.AttributeType) }, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.ValueInfoProto), global::Onnx.ValueInfoProto.Parser, new[]{ "Name", "Type", "DocString" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.NodeProto), global::Onnx.NodeProto.Parser, new[]{ "Input", "Output", "Name", "OpType", "Domain", "Attribute", "DocString" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.ModelProto), global::Onnx.ModelProto.Parser, new[]{ "IrVersion", "OpsetImport", "ProducerName", "ProducerVersion", "Domain", "ModelVersion", "DocString", "Graph", "Functions", "MetadataProps" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.StringStringEntryProto), global::Onnx.StringStringEntryProto.Parser, new[]{ "Key", "Value" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorAnnotation), global::Onnx.TensorAnnotation.Parser, new[]{ "TensorName", "QuantParameterTensorNames" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.GraphProto), global::Onnx.GraphProto.Parser, new[]{ "Node", "Name", "Initializer", "SparseInitializer", "DocString", "Input", "Output", "ValueInfo", "QuantizationAnnotation" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorProto), global::Onnx.TensorProto.Parser, new[]{ "Dims", "DataType", "Segment", "FloatData", "Int32Data", "StringData", "Int64Data", "Name", "DocString", "RawData", "ExternalData", "DataLocation", "DoubleData", "Uint64Data" }, null, new[]{ typeof(global::Onnx.TensorProto.Types.DataType), typeof(global::Onnx.TensorProto.Types.DataLocation) }, null, new pbr::GeneratedClrTypeInfo[] { new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorProto.Types.Segment), global::Onnx.TensorProto.Types.Segment.Parser, new[]{ "Begin", "End" }, null, null, null, null)}),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.SparseTensorProto), global::Onnx.SparseTensorProto.Parser, new[]{ "Values", "Indices", "Dims" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorShapeProto), global::Onnx.TensorShapeProto.Parser, new[]{ "Dim" }, null, null, null, new pbr::GeneratedClrTypeInfo[] { new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TensorShapeProto.Types.Dimension), global::Onnx.TensorShapeProto.Types.Dimension.Parser, new[]{ "DimValue", "DimParam", "Denotation" }, new[]{ "Value" }, null, null, null)}),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto), global::Onnx.TypeProto.Parser, new[]{ "TensorType", "SequenceType", "MapType", "SparseTensorType", "OpaqueType", "Denotation" }, new[]{ "Value" }, null, null, new pbr::GeneratedClrTypeInfo[] { new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Tensor), global::Onnx.TypeProto.Types.Tensor.Parser, new[]{ "ElemType", "Shape" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Sequence), global::Onnx.TypeProto.Types.Sequence.Parser, new[]{ "ElemType" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Map), global::Onnx.TypeProto.Types.Map.Parser, new[]{ "KeyType", "ValueType" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.SparseTensor), global::Onnx.TypeProto.Types.SparseTensor.Parser, new[]{ "ElemType", "Shape" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.TypeProto.Types.Opaque), global::Onnx.TypeProto.Types.Opaque.Parser, new[]{ "Domain", "Name" }, null, null, null, null)}),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.OperatorSetIdProto), global::Onnx.OperatorSetIdProto.Parser, new[]{ "Domain", "Version" }, null, null, null, null),
+            new pbr::GeneratedClrTypeInfo(typeof(global::Onnx.FunctionProto), global::Onnx.FunctionProto.Parser, new[]{ "Name", "SinceVersion", "Status", "Input", "Output", "Attribute", "Node", "DocString" }, null, null, null, null)
           }));
     }
     #endregion
@@ -304,7 +304,7 @@ public string DocString {
 
     /// <summary>Field number for the "type" field.</summary>
     public const int TypeFieldNumber = 20;
-    private global::Onnx.AttributeProto.Types.AttributeType type_ = 0;
+    private global::Onnx.AttributeProto.Types.AttributeType type_ = global::Onnx.AttributeProto.Types.AttributeType.Undefined;
     /// <summary>
     /// The type field MUST be present for this version of the IR.
     /// For 0.0.1 versions of the IR, this field was not defined, and
@@ -521,7 +521,7 @@ public override int GetHashCode() {
       if (Name.Length != 0) hash ^= Name.GetHashCode();
       if (RefAttrName.Length != 0) hash ^= RefAttrName.GetHashCode();
       if (DocString.Length != 0) hash ^= DocString.GetHashCode();
-      if (Type != 0) hash ^= Type.GetHashCode();
+      if (Type != global::Onnx.AttributeProto.Types.AttributeType.Undefined) hash ^= Type.GetHashCode();
       if (F != 0F) hash ^= pbc::ProtobufEqualityComparers.BitwiseSingleEqualityComparer.GetHashCode(F);
       if (I != 0L) hash ^= I.GetHashCode();
       if (S.Length != 0) hash ^= S.GetHashCode();
@@ -580,7 +580,7 @@ public void WriteTo(pb::CodedOutputStream output) {
         output.WriteRawTag(106);
         output.WriteString(DocString);
       }
-      if (Type != 0) {
+      if (Type != global::Onnx.AttributeProto.Types.AttributeType.Undefined) {
         output.WriteRawTag(160, 1);
         output.WriteEnum((int) Type);
       }
@@ -610,7 +610,7 @@ public int CalculateSize() {
       if (DocString.Length != 0) {
         size += 1 + pb::CodedOutputStream.ComputeStringSize(DocString);
       }
-      if (Type != 0) {
+      if (Type != global::Onnx.AttributeProto.Types.AttributeType.Undefined) {
         size += 2 + pb::CodedOutputStream.ComputeEnumSize((int) Type);
       }
       if (F != 0F) {
@@ -657,7 +657,7 @@ public void MergeFrom(AttributeProto other) {
       if (other.DocString.Length != 0) {
         DocString = other.DocString;
       }
-      if (other.Type != 0) {
+      if (other.Type != global::Onnx.AttributeProto.Types.AttributeType.Undefined) {
         Type = other.Type;
       }
       if (other.F != 0F) {
@@ -671,19 +671,19 @@ public void MergeFrom(AttributeProto other) {
       }
       if (other.t_ != null) {
         if (t_ == null) {
-          t_ = new global::Onnx.TensorProto();
+          T = new global::Onnx.TensorProto();
         }
         T.MergeFrom(other.T);
       }
       if (other.g_ != null) {
         if (g_ == null) {
-          g_ = new global::Onnx.GraphProto();
+          G = new global::Onnx.GraphProto();
         }
         G.MergeFrom(other.G);
       }
       if (other.sparseTensor_ != null) {
         if (sparseTensor_ == null) {
-          sparseTensor_ = new global::Onnx.SparseTensorProto();
+          SparseTensor = new global::Onnx.SparseTensorProto();
         }
         SparseTensor.MergeFrom(other.SparseTensor);
       }
@@ -722,16 +722,16 @@ public void MergeFrom(pb::CodedInputStream input) {
           }
           case 42: {
             if (t_ == null) {
-              t_ = new global::Onnx.TensorProto();
+              T = new global::Onnx.TensorProto();
             }
-            input.ReadMessage(t_);
+            input.ReadMessage(T);
             break;
           }
           case 50: {
             if (g_ == null) {
-              g_ = new global::Onnx.GraphProto();
+              G = new global::Onnx.GraphProto();
             }
-            input.ReadMessage(g_);
+            input.ReadMessage(G);
             break;
           }
           case 58:
@@ -761,7 +761,7 @@ public void MergeFrom(pb::CodedInputStream input) {
             break;
           }
           case 160: {
-            type_ = (global::Onnx.AttributeProto.Types.AttributeType) input.ReadEnum();
+            Type = (global::Onnx.AttributeProto.Types.AttributeType) input.ReadEnum();
             break;
           }
           case 170: {
@@ -770,9 +770,9 @@ public void MergeFrom(pb::CodedInputStream input) {
           }
           case 178: {
             if (sparseTensor_ == null) {
-              sparseTensor_ = new global::Onnx.SparseTensorProto();
+              SparseTensor = new global::Onnx.SparseTensorProto();
             }
-            input.ReadMessage(sparseTensor_);
+            input.ReadMessage(SparseTensor);
             break;
           }
           case 186: {
@@ -978,7 +978,7 @@ public void MergeFrom(ValueInfoProto other) {
       }
       if (other.type_ != null) {
         if (type_ == null) {
-          type_ = new global::Onnx.TypeProto();
+          Type = new global::Onnx.TypeProto();
         }
         Type.MergeFrom(other.Type);
       }
@@ -1002,9 +1002,9 @@ public void MergeFrom(pb::CodedInputStream input) {
           }
           case 18: {
             if (type_ == null) {
-              type_ = new global::Onnx.TypeProto();
+              Type = new global::Onnx.TypeProto();
             }
-            input.ReadMessage(type_);
+            input.ReadMessage(Type);
             break;
           }
           case 26: {
@@ -1673,7 +1673,7 @@ public void MergeFrom(ModelProto other) {
       }
       if (other.graph_ != null) {
         if (graph_ == null) {
-          graph_ = new global::Onnx.GraphProto();
+          Graph = new global::Onnx.GraphProto();
         }
         Graph.MergeFrom(other.Graph);
       }
@@ -1716,9 +1716,9 @@ public void MergeFrom(pb::CodedInputStream input) {
           }
           case 58: {
             if (graph_ == null) {
-              graph_ = new global::Onnx.GraphProto();
+              Graph = new global::Onnx.GraphProto();
             }
-            input.ReadMessage(graph_);
+            input.ReadMessage(Graph);
             break;
           }
           case 66: {
@@ -2627,7 +2627,7 @@ public string DocString {
 
     /// <summary>Field number for the "data_location" field.</summary>
     public const int DataLocationFieldNumber = 14;
-    private global::Onnx.TensorProto.Types.DataLocation dataLocation_ = 0;
+    private global::Onnx.TensorProto.Types.DataLocation dataLocation_ = global::Onnx.TensorProto.Types.DataLocation.Default;
     /// <summary>
     /// If value not set, data is stored in raw_data (if set) otherwise in type-specified field.
     /// </summary>
@@ -2717,7 +2717,7 @@ public override int GetHashCode() {
       if (DocString.Length != 0) hash ^= DocString.GetHashCode();
       if (RawData.Length != 0) hash ^= RawData.GetHashCode();
       hash ^= externalData_.GetHashCode();
-      if (DataLocation != 0) hash ^= DataLocation.GetHashCode();
+      if (DataLocation != global::Onnx.TensorProto.Types.DataLocation.Default) hash ^= DataLocation.GetHashCode();
       hash ^= doubleData_.GetHashCode();
       hash ^= uint64Data_.GetHashCode();
       if (_unknownFields != null) {
@@ -2761,7 +2761,7 @@ public void WriteTo(pb::CodedOutputStream output) {
         output.WriteString(DocString);
       }
       externalData_.WriteTo(output, _repeated_externalData_codec);
-      if (DataLocation != 0) {
+      if (DataLocation != global::Onnx.TensorProto.Types.DataLocation.Default) {
         output.WriteRawTag(112);
         output.WriteEnum((int) DataLocation);
       }
@@ -2794,7 +2794,7 @@ public int CalculateSize() {
         size += 1 + pb::CodedOutputStream.ComputeBytesSize(RawData);
       }
       size += externalData_.CalculateSize(_repeated_externalData_codec);
-      if (DataLocation != 0) {
+      if (DataLocation != global::Onnx.TensorProto.Types.DataLocation.Default) {
         size += 1 + pb::CodedOutputStream.ComputeEnumSize((int) DataLocation);
       }
       size += doubleData_.CalculateSize(_repeated_doubleData_codec);
@@ -2816,7 +2816,7 @@ public void MergeFrom(TensorProto other) {
       }
       if (other.segment_ != null) {
         if (segment_ == null) {
-          segment_ = new global::Onnx.TensorProto.Types.Segment();
+          Segment = new global::Onnx.TensorProto.Types.Segment();
         }
         Segment.MergeFrom(other.Segment);
       }
@@ -2834,7 +2834,7 @@ public void MergeFrom(TensorProto other) {
         RawData = other.RawData;
       }
       externalData_.Add(other.externalData_);
-      if (other.DataLocation != 0) {
+      if (other.DataLocation != global::Onnx.TensorProto.Types.DataLocation.Default) {
         DataLocation = other.DataLocation;
       }
       doubleData_.Add(other.doubleData_);
@@ -2861,9 +2861,9 @@ public void MergeFrom(pb::CodedInputStream input) {
           }
           case 26: {
             if (segment_ == null) {
-              segment_ = new global::Onnx.TensorProto.Types.Segment();
+              Segment = new global::Onnx.TensorProto.Types.Segment();
             }
-            input.ReadMessage(segment_);
+            input.ReadMessage(Segment);
             break;
           }
           case 34:
@@ -2912,7 +2912,7 @@ public void MergeFrom(pb::CodedInputStream input) {
             break;
           }
           case 112: {
-            dataLocation_ = (global::Onnx.TensorProto.Types.DataLocation) input.ReadEnum();
+            DataLocation = (global::Onnx.TensorProto.Types.DataLocation) input.ReadEnum();
             break;
           }
         }
@@ -3327,13 +3327,13 @@ public void MergeFrom(SparseTensorProto other) {
       }
       if (other.values_ != null) {
         if (values_ == null) {
-          values_ = new global::Onnx.TensorProto();
+          Values = new global::Onnx.TensorProto();
         }
         Values.MergeFrom(other.Values);
       }
       if (other.indices_ != null) {
         if (indices_ == null) {
-          indices_ = new global::Onnx.TensorProto();
+          Indices = new global::Onnx.TensorProto();
         }
         Indices.MergeFrom(other.Indices);
       }
@@ -3351,16 +3351,16 @@ public void MergeFrom(pb::CodedInputStream input) {
             break;
           case 10: {
             if (values_ == null) {
-              values_ = new global::Onnx.TensorProto();
+              Values = new global::Onnx.TensorProto();
             }
-            input.ReadMessage(values_);
+            input.ReadMessage(Values);
             break;
           }
           case 18: {
             if (indices_ == null) {
-              indices_ = new global::Onnx.TensorProto();
+              Indices = new global::Onnx.TensorProto();
             }
-            input.ReadMessage(indices_);
+            input.ReadMessage(Indices);
             break;
           }
           case 26:
@@ -4240,7 +4240,7 @@ public void MergeFrom(Tensor other) {
           }
           if (other.shape_ != null) {
             if (shape_ == null) {
-              shape_ = new global::Onnx.TensorShapeProto();
+              Shape = new global::Onnx.TensorShapeProto();
             }
             Shape.MergeFrom(other.Shape);
           }
@@ -4261,9 +4261,9 @@ public void MergeFrom(pb::CodedInputStream input) {
               }
               case 18: {
                 if (shape_ == null) {
-                  shape_ = new global::Onnx.TensorShapeProto();
+                  Shape = new global::Onnx.TensorShapeProto();
                 }
-                input.ReadMessage(shape_);
+                input.ReadMessage(Shape);
                 break;
               }
             }
@@ -4386,7 +4386,7 @@ public void MergeFrom(Sequence other) {
           }
           if (other.elemType_ != null) {
             if (elemType_ == null) {
-              elemType_ = new global::Onnx.TypeProto();
+              ElemType = new global::Onnx.TypeProto();
             }
             ElemType.MergeFrom(other.ElemType);
           }
@@ -4403,9 +4403,9 @@ public void MergeFrom(pb::CodedInputStream input) {
                 break;
               case 10: {
                 if (elemType_ == null) {
-                  elemType_ = new global::Onnx.TypeProto();
+                  ElemType = new global::Onnx.TypeProto();
                 }
-                input.ReadMessage(elemType_);
+                input.ReadMessage(ElemType);
                 break;
               }
             }
@@ -4556,7 +4556,7 @@ public void MergeFrom(Map other) {
           }
           if (other.valueType_ != null) {
             if (valueType_ == null) {
-              valueType_ = new global::Onnx.TypeProto();
+              ValueType = new global::Onnx.TypeProto();
             }
             ValueType.MergeFrom(other.ValueType);
           }
@@ -4577,9 +4577,9 @@ public void MergeFrom(pb::CodedInputStream input) {
               }
               case 18: {
                 if (valueType_ == null) {
-                  valueType_ = new global::Onnx.TypeProto();
+                  ValueType = new global::Onnx.TypeProto();
                 }
-                input.ReadMessage(valueType_);
+                input.ReadMessage(ValueType);
                 break;
               }
             }
@@ -4724,7 +4724,7 @@ public void MergeFrom(SparseTensor other) {
           }
           if (other.shape_ != null) {
             if (shape_ == null) {
-              shape_ = new global::Onnx.TensorShapeProto();
+              Shape = new global::Onnx.TensorShapeProto();
             }
             Shape.MergeFrom(other.Shape);
           }
@@ -4745,9 +4745,9 @@ public void MergeFrom(pb::CodedInputStream input) {
               }
               case 18: {
                 if (shape_ == null) {
-                  shape_ = new global::Onnx.TensorShapeProto();
+                  Shape = new global::Onnx.TensorShapeProto();
                 }
-                input.ReadMessage(shape_);
+                input.ReadMessage(Shape);
                 break;
               }
             }
@@ -5170,7 +5170,7 @@ public long SinceVersion {
 
     /// <summary>Field number for the "status" field.</summary>
     public const int StatusFieldNumber = 3;
-    private global::Onnx.OperatorStatus status_ = 0;
+    private global::Onnx.OperatorStatus status_ = global::Onnx.OperatorStatus.Experimental;
     /// <summary>
     /// This field indicates whether the syntax, semantics, or presence
     /// of this function is in an experimental or stable stage. Once an
@@ -5281,7 +5281,7 @@ public override int GetHashCode() {
       int hash = 1;
       if (Name.Length != 0) hash ^= Name.GetHashCode();
       if (SinceVersion != 0L) hash ^= SinceVersion.GetHashCode();
-      if (Status != 0) hash ^= Status.GetHashCode();
+      if (Status != global::Onnx.OperatorStatus.Experimental) hash ^= Status.GetHashCode();
       hash ^= input_.GetHashCode();
       hash ^= output_.GetHashCode();
       hash ^= attribute_.GetHashCode();
@@ -5308,7 +5308,7 @@ public void WriteTo(pb::CodedOutputStream output) {
         output.WriteRawTag(16);
         output.WriteInt64(SinceVersion);
       }
-      if (Status != 0) {
+      if (Status != global::Onnx.OperatorStatus.Experimental) {
         output.WriteRawTag(24);
         output.WriteEnum((int) Status);
       }
@@ -5334,7 +5334,7 @@ public int CalculateSize() {
       if (SinceVersion != 0L) {
         size += 1 + pb::CodedOutputStream.ComputeInt64Size(SinceVersion);
       }
-      if (Status != 0) {
+      if (Status != global::Onnx.OperatorStatus.Experimental) {
         size += 1 + pb::CodedOutputStream.ComputeEnumSize((int) Status);
       }
       size += input_.CalculateSize(_repeated_input_codec);
@@ -5361,7 +5361,7 @@ public void MergeFrom(FunctionProto other) {
       if (other.SinceVersion != 0L) {
         SinceVersion = other.SinceVersion;
       }
-      if (other.Status != 0) {
+      if (other.Status != global::Onnx.OperatorStatus.Experimental) {
         Status = other.Status;
       }
       input_.Add(other.input_);
@@ -5391,7 +5391,7 @@ public void MergeFrom(pb::CodedInputStream input) {
             break;
           }
           case 24: {
-            status_ = (global::Onnx.OperatorStatus) input.ReadEnum();
+            Status = (global::Onnx.OperatorStatus) input.ReadEnum();
             break;
           }
           case 34: {
diff --git a/onnxruntime/core/framework/path_lib.h b/onnxruntime/core/framework/path_lib.h
index 89cfaef4188e7..1339450c9a664 100644
--- a/onnxruntime/core/framework/path_lib.h
+++ b/onnxruntime/core/framework/path_lib.h
@@ -34,11 +34,14 @@ ptrdiff_t OrtStrToPtrDiff(const T* nptr, T** endptr);
 template <>
 inline ptrdiff_t OrtStrToPtrDiff<wchar_t>(const wchar_t* nptr, wchar_t** endptr) {
 #ifdef _WIN32
-  // TODO: on x86_32, it should be wcstol
+#ifdef _M_AMD64
   return _wcstoi64(nptr, endptr, 10);
 #else
   return wcstol(nptr, endptr, 10);
 #endif
+#else
+  return wcstol(nptr, endptr, 10);
+#endif
 }
 
 template <typename T>
@@ -57,11 +60,14 @@ inline size_t OrtStrftime<wchar_t>(wchar_t* strDest, size_t maxsize, const wchar
 template <>
 inline ptrdiff_t OrtStrToPtrDiff<char>(const char* nptr, char** endptr) {
 #ifdef _WIN32
-  // TODO: on x86_32, it should be strtol
+#ifdef _M_AMD64
   return _strtoi64(nptr, endptr, 10);
 #else
   return strtol(nptr, endptr, 10);
 #endif
+#else
+  return strtol(nptr, endptr, 10);
+#endif
 }
 
 template <>
diff --git a/onnxruntime/core/graph/model.cc b/onnxruntime/core/graph/model.cc
index f8711e0c91481..a91d5a7772396 100644
--- a/onnxruntime/core/graph/model.cc
+++ b/onnxruntime/core/graph/model.cc
@@ -428,7 +428,7 @@ Status Model::Load(int fd, ONNX_NAMESPACE::ModelProto& model_proto) {
   CodedInputStream cis(&fs);
 
   // Allows protobuf library versions < 3.2.0 to parse messages greater than 64MB.
-  cis.SetTotalBytesLimit(INT_MAX, INT_MAX);
+  cis.SetTotalBytesLimit(INT_MAX);
   if (!model_proto->ParseFromCodedStream(&cis)) {
     return Status(ONNXRUNTIME, INVALID_PROTOBUF, "Protobuf parsing failed.");
   }
diff --git a/onnxruntime/core/session/onnxruntime_c_api.cc b/onnxruntime/core/session/onnxruntime_c_api.cc
index 983889f93a561..3a3dab335bcfa 100644
--- a/onnxruntime/core/session/onnxruntime_c_api.cc
+++ b/onnxruntime/core/session/onnxruntime_c_api.cc
@@ -245,7 +245,7 @@ OrtStatus* CreateTensorImpl(MLDataType ml_type, const int64_t* shape, size_t sha
   size_t elem_count = 1;
   std::vector<int64_t> shapes(shape_len);
   for (size_t i = 0; i != shape_len; ++i) {
-    elem_count *= shape[i];
+    elem_count *= static_cast<size_t>(shape[i]);
     shapes[i] = shape[i];
   }
 
@@ -971,28 +971,31 @@ static OrtStatus* OrtGetValueImplMapHelper(const OrtValue* p_ml_value, int index
   using TVal = typename T::mapped_type;
   auto& data = p_ml_value->Get<T>();
   int64_t num_kv_pairs = data.size();
+#if defined(_WIN32) && !defined(_M_AMD64)
+  ORT_ENFORCE(static_cast<uint64_t>(num_kv_pairs) < std::numeric_limits<size_t>::max());
+#endif
   switch (index) {
     case 0: {  // user is requesting keys
       std::vector<TKey> vec;
-      vec.reserve(num_kv_pairs);
+      vec.reserve(static_cast<size_t>(num_kv_pairs));
       for (const auto& kv : data) {
         vec.push_back(kv.first);
       }
       std::vector<int64_t> dims{num_kv_pairs};
       OrtStatus* st = OrtApis::CreateTensorAsOrtValue(allocator, dims.data(), dims.size(),
                                                       GetONNXTensorElementDataType<TKey>(), out);
-      return st ? st : PopulateTensorWithData(*out, vec.data(), num_kv_pairs, sizeof(TKey));
+      return st ? st : PopulateTensorWithData(*out, vec.data(), static_cast<size_t>(num_kv_pairs), sizeof(TKey));
     }
     case 1: {  // user is requesting values
       std::vector<TVal> vec;
-      vec.reserve(num_kv_pairs);
+      vec.reserve(static_cast<size_t>(num_kv_pairs));
       for (const auto& kv : data) {
         vec.push_back(kv.second);
       }
       std::vector<int64_t> dims{num_kv_pairs};
       OrtStatus* st = OrtApis::CreateTensorAsOrtValue(allocator, dims.data(), dims.size(),
                                                       GetONNXTensorElementDataType<TVal>(), out);
-      return st ? st : PopulateTensorWithData(*out, vec.data(), num_kv_pairs, sizeof(TVal));
+      return st ? st : PopulateTensorWithData(*out, vec.data(), static_cast<size_t>(num_kv_pairs), sizeof(TVal));
     }
     default:
       return OrtApis::CreateStatus(ORT_FAIL, "Invalid index requested for map type.");
@@ -1200,7 +1203,9 @@ static OrtStatus* OrtCreateMapMLValue(const Tensor& key_tensor, const Tensor& va
   // iterate through the key and value tensors and populate map
   auto key_data = key_tensor.Data<KeyType>();
   auto value_data = value_tensor.Data<ValueType>();
-  size_t num_kv_pairs = key_tensor.Shape().Size();
+  auto len = key_tensor.Shape().Size();
+  ORT_ENFORCE(len >= 0 && static_cast<uint64_t>(len) < std::numeric_limits<size_t>::max());
+  size_t num_kv_pairs = static_cast<size_t>(key_tensor.Shape().Size());
   for (size_t n = 0; n < num_kv_pairs; ++n, ++key_data, ++value_data) {
     map_ptr->insert({*key_data, *value_data});
   }
diff --git a/onnxruntime/core/util/protobuf_parsing_utils.cc b/onnxruntime/core/util/protobuf_parsing_utils.cc
deleted file mode 100644
index 7cdf60bb783a3..0000000000000
--- a/onnxruntime/core/util/protobuf_parsing_utils.cc
+++ /dev/null
@@ -1,481 +0,0 @@
-// Protocol Buffers - Google's data interchange format
-// Copyright 2008 Google Inc.  All rights reserved.
-// https://developers.google.com/protocol-buffers/
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-// Author: kenton@google.com (Kenton Varda)
-//  Based on original Protocol Buffers design by
-//  Sanjay Ghemawat, Jeff Dean, and others.
-
-// Modifications Copyright (c) Microsoft.
-// Introduced this file based on suggestion by Kenton Varda as seen here
-// https://groups.google.com/forum/#!topic/protobuf/J3H0GSlrfIo
-// The idea is to retain the same efficiency of loading models.
-
-#ifndef _MSC_VER
-#include <unistd.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#endif
-#include <errno.h>
-#include <iostream>
-#include <algorithm>
-#include <google/protobuf/stubs/common.h>
-#include <google/protobuf/stubs/logging.h>
-#include <google/protobuf/stubs/stl_util.h>
-#ifdef _WIN32
-#include <google/protobuf/stubs/io_win32.h>
-#endif
-#include "protobuf_parsing_utils.h"
-#include <string.h>
-
-namespace google {
-namespace protobuf {
-namespace io {
-
-#ifdef _WIN32
-// Win32 lseek is broken:  If invoked on a non-seekable file descriptor, its
-// return value is undefined.  We re-define it to always produce an error.
-#define lseek(fd, offset, origin) ((off_t)-1)
-// DO NOT include <io.h>, instead create functions in io_win32.{h,cc} and import
-// them like we do below.
-using google::protobuf::internal::win32::access;
-using google::protobuf::internal::win32::close;
-using google::protobuf::internal::win32::open;
-using google::protobuf::internal::win32::read;
-using google::protobuf::internal::win32::write;
-#endif
-
-namespace {
-
-// EINTR sucks.
-int close_no_eintr(int fd) {
-  int result;
-  do {
-    result = close(fd);
-  } while (result < 0 && errno == EINTR);
-  return result;
-}
-
-}  // namespace
-
-// ===================================================================
-
-FileInputStream::FileInputStream(int file_descriptor, int block_size)
-    : copying_input_(file_descriptor),
-      impl_(&copying_input_, block_size) {
-}
-
-bool FileInputStream::Close() {
-  return copying_input_.Close();
-}
-
-bool FileInputStream::Next(const void** data, int* size) {
-  return impl_.Next(data, size);
-}
-
-void FileInputStream::BackUp(int count) {
-  impl_.BackUp(count);
-}
-
-bool FileInputStream::Skip(int count) {
-  return impl_.Skip(count);
-}
-
-int64 FileInputStream::ByteCount() const {
-  return impl_.ByteCount();
-}
-
-FileInputStream::CopyingFileInputStream::CopyingFileInputStream(
-    int file_descriptor)
-    : file_(file_descriptor),
-      close_on_delete_(false),
-      is_closed_(false),
-      errno_(0),
-      previous_seek_failed_(false) {
-}
-
-FileInputStream::CopyingFileInputStream::~CopyingFileInputStream() {
-  if (close_on_delete_) {
-    if (!Close()) {
-#ifdef _WIN32
-      const size_t errmsglen = 100;  // strerrorlen_s not available on MSVC, https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/strerror-s-strerror-s-wcserror-s-wcserror-s?view=vs-2017
-      char errmsg[errmsglen];
-      strerror_s(errmsg, errmsglen, errno);
-      GOOGLE_LOG(ERROR) << "close() failed: " << errmsg;
-#else  // POSIX
-      char* e_msg = strerror(errno);
-      GOOGLE_LOG(ERROR) << "close() failed: " << e_msg;
-#endif
-    }
-  }
-}
-
-bool FileInputStream::CopyingFileInputStream::Close() {
-  GOOGLE_CHECK(!is_closed_);
-
-  is_closed_ = true;
-  if (close_no_eintr(file_) != 0) {
-    // The docs on close() do not specify whether a file descriptor is still
-    // open after close() fails with EIO.  However, the glibc source code
-    // seems to indicate that it is not.
-    errno_ = errno;
-    return false;
-  }
-
-  return true;
-}
-
-int FileInputStream::CopyingFileInputStream::Read(void* buffer, int size) {
-  GOOGLE_CHECK(!is_closed_);
-
-  int result;
-  do {
-    result = read(file_, buffer, size);
-  } while (result < 0 && errno == EINTR);
-
-  if (result < 0) {
-    // Read error (not EOF).
-    errno_ = errno;
-  }
-
-  return result;
-}
-
-int FileInputStream::CopyingFileInputStream::Skip(int count) {
-  GOOGLE_CHECK(!is_closed_);
-
-  if (!previous_seek_failed_ && lseek(file_, count, SEEK_CUR) != static_cast<off_t>(-1)) {
-    // Seek succeeded.
-    return count;
-  }
-    // Failed to seek.
-
-    // Note to self:  Don't seek again.  This file descriptor doesn't
-    // support it.
-    previous_seek_failed_ = true;
-
-    // Use the default implementation.
-    return CopyingInputStream::Skip(count);
-}
-
-// ===================================================================
-
-FileOutputStream::FileOutputStream(int file_descriptor, int block_size)
-    : copying_output_(file_descriptor),
-      impl_(&copying_output_, block_size) {
-}
-
-FileOutputStream::~FileOutputStream() {
-  impl_.Flush();
-}
-
-bool FileOutputStream::Close() {
-  bool flush_succeeded = impl_.Flush();
-  return copying_output_.Close() && flush_succeeded;
-}
-
-bool FileOutputStream::Flush() {
-  return impl_.Flush();
-}
-
-bool FileOutputStream::Next(void** data, int* size) {
-  return impl_.Next(data, size);
-}
-
-void FileOutputStream::BackUp(int count) {
-  impl_.BackUp(count);
-}
-
-int64 FileOutputStream::ByteCount() const {
-  return impl_.ByteCount();
-}
-
-FileOutputStream::CopyingFileOutputStream::CopyingFileOutputStream(
-    int file_descriptor)
-    : file_(file_descriptor),
-      close_on_delete_(false),
-      is_closed_(false),
-      errno_(0) {
-}
-
-FileOutputStream::CopyingFileOutputStream::~CopyingFileOutputStream() {
-  if (close_on_delete_) {
-    if (!Close()) {
-#ifdef _WIN32
-      // https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/strerror-s-strerror-s-wcserror-s-wcserror-s?view=vs-2017
-      const size_t errmsglen = 100;  //strerrorlen_s not available on MSVC
-      char errmsg[errmsglen];
-      strerror_s(errmsg, errmsglen, errno);
-      GOOGLE_LOG(ERROR) << "close() failed: " << errmsg;
-#else  // POSIX
-      char* e_msg = strerror(errno);
-      GOOGLE_LOG(ERROR) << "close() failed: " << e_msg;
-#endif
-    }
-  }
-}
-
-bool FileOutputStream::CopyingFileOutputStream::Close() {
-  GOOGLE_CHECK(!is_closed_);
-
-  is_closed_ = true;
-  if (close_no_eintr(file_) != 0) {
-    // The docs on close() do not specify whether a file descriptor is still
-    // open after close() fails with EIO.  However, the glibc source code
-    // seems to indicate that it is not.
-    errno_ = errno;
-    return false;
-  }
-
-  return true;
-}
-
-bool FileOutputStream::CopyingFileOutputStream::Write(
-    const void* buffer, int size) {
-  GOOGLE_CHECK(!is_closed_);
-  int total_written = 0;
-
-  const uint8* buffer_base = reinterpret_cast<const uint8*>(buffer);
-
-  while (total_written < size) {
-    int bytes;
-    do {
-      bytes = write(file_, buffer_base + total_written, size - total_written);
-    } while (bytes < 0 && errno == EINTR);
-
-    if (bytes <= 0) {
-      // Write error.
-
-      // FIXME(kenton):  According to the man page, if write() returns zero,
-      //   there was no error; write() simply did not write anything.  It's
-      //   unclear under what circumstances this might happen, but presumably
-      //   errno won't be set in this case.  I am confused as to how such an
-      //   event should be handled.  For now I'm treating it as an error, since
-      //   retrying seems like it could lead to an infinite loop.  I suspect
-      //   this never actually happens anyway.
-
-      if (bytes < 0) {
-        errno_ = errno;
-      }
-      return false;
-    }
-    total_written += bytes;
-  }
-
-  return true;
-}
-
-// ===================================================================
-
-IstreamInputStream::IstreamInputStream(std::istream* input, int block_size)
-    : copying_input_(input), impl_(&copying_input_, block_size) {}
-
-bool IstreamInputStream::Next(const void** data, int* size) {
-  return impl_.Next(data, size);
-}
-
-void IstreamInputStream::BackUp(int count) {
-  impl_.BackUp(count);
-}
-
-bool IstreamInputStream::Skip(int count) {
-  return impl_.Skip(count);
-}
-
-int64 IstreamInputStream::ByteCount() const {
-  return impl_.ByteCount();
-}
-
-IstreamInputStream::CopyingIstreamInputStream::CopyingIstreamInputStream(
-    std::istream* input)
-    : input_(input) {}
-
-IstreamInputStream::CopyingIstreamInputStream::~CopyingIstreamInputStream() = default;
-
-int IstreamInputStream::CopyingIstreamInputStream::Read(
-    void* buffer, int size) {
-  input_->read(reinterpret_cast<char*>(buffer), size);
-  int result = static_cast<int>(input_->gcount());
-  if (result == 0 && input_->fail() && !input_->eof()) {
-    return -1;
-  }
-  return result;
-}
-
-// ===================================================================
-
-OstreamOutputStream::OstreamOutputStream(std::ostream* output, int block_size)
-    : copying_output_(output), impl_(&copying_output_, block_size) {}
-
-OstreamOutputStream::~OstreamOutputStream() {
-  impl_.Flush();
-}
-
-bool OstreamOutputStream::Next(void** data, int* size) {
-  return impl_.Next(data, size);
-}
-
-void OstreamOutputStream::BackUp(int count) {
-  impl_.BackUp(count);
-}
-
-int64 OstreamOutputStream::ByteCount() const {
-  return impl_.ByteCount();
-}
-
-OstreamOutputStream::CopyingOstreamOutputStream::CopyingOstreamOutputStream(
-    std::ostream* output)
-    : output_(output) {}
-
-OstreamOutputStream::CopyingOstreamOutputStream::~CopyingOstreamOutputStream() = default;
-
-bool OstreamOutputStream::CopyingOstreamOutputStream::Write(
-    const void* buffer, int size) {
-  output_->write(reinterpret_cast<const char*>(buffer), size);
-  return output_->good();
-}
-
-// ===================================================================
-
-ConcatenatingInputStream::ConcatenatingInputStream(
-    ZeroCopyInputStream* const streams[], int count)
-    : streams_(streams), stream_count_(count), bytes_retired_(0) {
-}
-
-bool ConcatenatingInputStream::Next(const void** data, int* size) {
-  while (stream_count_ > 0) {
-    if (streams_[0]->Next(data, size)) return true;
-
-    // That stream is done.  Advance to the next one.
-    bytes_retired_ += streams_[0]->ByteCount();
-    ++streams_;
-    --stream_count_;
-  }
-
-  // No more streams.
-  return false;
-}
-
-void ConcatenatingInputStream::BackUp(int count) {
-  if (stream_count_ > 0) {
-    streams_[0]->BackUp(count);
-  } else {
-    GOOGLE_LOG(DFATAL) << "Can't BackUp() after failed Next().";
-  }
-}
-
-bool ConcatenatingInputStream::Skip(int count) {
-  while (stream_count_ > 0) {
-    // Assume that ByteCount() can be used to find out how much we actually
-    // skipped when Skip() fails.
-    int64 target_byte_count = streams_[0]->ByteCount() + count;
-    if (streams_[0]->Skip(count)) return true;
-
-    // Hit the end of the stream.  Figure out how many more bytes we still have
-    // to skip.
-    int64 final_byte_count = streams_[0]->ByteCount();
-    GOOGLE_DCHECK_LT(final_byte_count, target_byte_count);
-    count = static_cast<int>(target_byte_count - final_byte_count);
-
-    // That stream is done.  Advance to the next one.
-    bytes_retired_ += final_byte_count;
-    ++streams_;
-    --stream_count_;
-  }
-
-  return false;
-}
-
-int64 ConcatenatingInputStream::ByteCount() const {
-  if (stream_count_ == 0) {
-    return bytes_retired_;
-  }
-  return bytes_retired_ + streams_[0]->ByteCount();
-}
-
-// ===================================================================
-
-LimitingInputStream::LimitingInputStream(ZeroCopyInputStream* input,
-                                         int64 limit)
-    : input_(input), limit_(limit) {
-  prior_bytes_read_ = input_->ByteCount();
-}
-
-LimitingInputStream::~LimitingInputStream() {
-  // If we overshot the limit, back up.
-  if (limit_ < 0) input_->BackUp(static_cast<int>(-limit_));
-}
-
-bool LimitingInputStream::Next(const void** data, int* size) {
-  if (limit_ <= 0) return false;
-  if (!input_->Next(data, size)) return false;
-
-  limit_ -= *size;
-  if (limit_ < 0) {
-    // We overshot the limit.  Reduce *size to hide the rest of the buffer.
-    *size += static_cast<int>(limit_);
-  }
-  return true;
-}
-
-void LimitingInputStream::BackUp(int count) {
-  if (limit_ < 0) {
-    input_->BackUp(static_cast<int>(count - limit_));
-    limit_ = count;
-  } else {
-    input_->BackUp(count);
-    limit_ += count;
-  }
-}
-
-bool LimitingInputStream::Skip(int count) {
-  if (count > limit_) {
-    if (limit_ < 0) return false;
-    input_->Skip(static_cast<int>(limit_));
-    limit_ = 0;
-    return false;
-  }
-    if (!input_->Skip(count)) return false;
-    limit_ -= count;
-    return true;
-}
-
-int64 LimitingInputStream::ByteCount() const {
-  if (limit_ < 0) {
-    return input_->ByteCount() + limit_ - prior_bytes_read_;
-  }
-  return input_->ByteCount() - prior_bytes_read_;
-}
-
-// ===================================================================
-
-}  // namespace io
-}  // namespace protobuf
-}  // namespace google
diff --git a/onnxruntime/core/util/protobuf_parsing_utils.h b/onnxruntime/core/util/protobuf_parsing_utils.h
index 34c261b79f41e..4c97562653f04 100644
--- a/onnxruntime/core/util/protobuf_parsing_utils.h
+++ b/onnxruntime/core/util/protobuf_parsing_utils.h
@@ -1,41 +1,3 @@
-// Protocol Buffers - Google's data interchange format
-// Copyright 2008 Google Inc.  All rights reserved.
-// https://developers.google.com/protocol-buffers/
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-// Author: kenton@google.com (Kenton Varda)
-//  Based on original Protocol Buffers design by
-//  Sanjay Ghemawat, Jeff Dean, and others.
-//
-// This file contains common implementations of the interfaces defined in
-// zero_copy_stream.h which are only included in the full (non-lite)
-// protobuf library.  These implementations include Unix file descriptors
-// and C++ iostreams.  See also:  zero_copy_stream_impl_lite.h
 
 #pragma once
 
@@ -317,30 +279,6 @@ class LIBPROTOBUF_EXPORT ConcatenatingInputStream : public ZeroCopyInputStream {
   GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(ConcatenatingInputStream);
 };
 
-// ===================================================================
-
-// A ZeroCopyInputStream which wraps some other stream and limits it to
-// a particular byte count.
-class LIBPROTOBUF_EXPORT LimitingInputStream : public ZeroCopyInputStream {
- public:
-  LimitingInputStream(ZeroCopyInputStream* input, int64 limit);
-  ~LimitingInputStream() override;
-
-  // implements ZeroCopyInputStream ----------------------------------
-  bool Next(const void** data, int* size) override;
-  void BackUp(int count) override;
-  bool Skip(int count) override;
-  int64 ByteCount() const override;
-
- private:
-  ZeroCopyInputStream* input_;
-  int64 limit_;             // Decreases as we go, becomes negative if we overshoot.
-  int64 prior_bytes_read_;  // Bytes read on underlying stream at construction
-
-  GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(LimitingInputStream);
-};
-
-// ===================================================================
 
 }  // namespace io
 }  // namespace protobuf
diff --git a/onnxruntime/test/ir/onnx_model_test.cc b/onnxruntime/test/ir/onnx_model_test.cc
index 77b45d7261f99..9a6da3a52e607 100644
--- a/onnxruntime/test/ir/onnx_model_test.cc
+++ b/onnxruntime/test/ir/onnx_model_test.cc
@@ -74,7 +74,7 @@ TEST(ONNXModelsTest1, bvlc_alexnet_1) {
   std::unique_ptr<ZeroCopyInputStream> raw_input(new FileInputStream(fd));
   std::unique_ptr<CodedInputStream> coded_input(new CodedInputStream(raw_input.get()));
   // Allows protobuf library versions < 3.2.0 to parse messages greater than 64MB.
-  coded_input->SetTotalBytesLimit(INT_MAX, INT_MAX);
+  coded_input->SetTotalBytesLimit(INT_MAX);
   ModelProto model_proto;
   bool result = model_proto.ParseFromCodedStream(coded_input.get());
   coded_input.reset();
@@ -124,7 +124,7 @@ TEST_P(ONNXModelsTest, LoadFromProtobuf) {
   ASSERT_TRUE(fd > 0);
   std::unique_ptr<ZeroCopyInputStream> raw_input(new FileInputStream(fd));
   std::unique_ptr<CodedInputStream> coded_input(new CodedInputStream(raw_input.get()));
-  coded_input->SetTotalBytesLimit(INT_MAX, INT_MAX);
+  coded_input->SetTotalBytesLimit(INT_MAX);
   std::unique_ptr<ModelProto> model_proto = onnxruntime::make_unique<ModelProto>();
   bool result = model_proto->ParseFromCodedStream(coded_input.get());
   coded_input.reset();
diff --git a/onnxruntime/test/onnx/TestCase.cc b/onnxruntime/test/onnx/TestCase.cc
index de09b17df8dcd..9b6e661c8cc50 100644
--- a/onnxruntime/test/onnx/TestCase.cc
+++ b/onnxruntime/test/onnx/TestCase.cc
@@ -203,12 +203,13 @@ class OnnxModelInfo : public TestModelInfo {
     if (!st.IsOK()) {
       ORT_THROW(st.ErrorMessage());
     }
-    google::protobuf::io::FileInputStream f(model_fd);
-    f.SetCloseOnDelete(true);
+
     ONNX_NAMESPACE::ModelProto model_pb;
-    if (!model_pb.ParseFromZeroCopyStream(&f)) {
+    if (!model_pb.ParseFromFileDescriptor(model_fd)) {
+      (void)Env::Default().FileClose(model_fd);
       ORT_THROW("Failed to load model because protobuf parsing failed.");
     }
+    (void)Env::Default().FileClose(model_fd);
 #ifdef __GNUG__
     const RE2::Anchor re2_anchor = RE2::UNANCHORED;
     re2::StringPiece text(model_url);
@@ -233,6 +234,7 @@ class OnnxModelInfo : public TestModelInfo {
       if (!init.has_name()) continue;
       initializer_names.insert(init.name());
     }
+    //Ignore the inputs that are already in initializers
     for (const auto& p : graph.input()) {
       if (!p.has_name()) ORT_THROW("input without name??");
       if (initializer_names.find(p.name()) == initializer_names.end()) input_value_info_.push_back(p);
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index d8bd447e099aa..6fa0b30456d03 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -354,14 +354,16 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
       sf.SetGraphOptimizationLevel(graph_optimization_level);
     }
 
-    static const char* cuda_flaky_tests[] = {"fp16_inception_v1", "fp16_shufflenet", "fp16_tiny_yolov2"};
-    static const char* dml_disabled_tests[] = {"mlperf_ssd_resnet34_1200", "mlperf_ssd_mobilenet_300", "mask_rcnn_keras", "mask_rcnn", "faster_rcnn"};
-    static const char* dnnl_disabled_tests[] = {"test_densenet121", "test_resnet18v2", "test_resnet34v2", "test_resnet50v2", "test_resnet101v2",
-                                                "test_resnet101v2", "test_vgg19", "tf_inception_resnet_v2", "tf_inception_v1", "tf_inception_v3", "tf_inception_v4", "tf_mobilenet_v1_1.0_224",
-                                                "tf_mobilenet_v2_1.0_224", "tf_mobilenet_v2_1.4_224", "tf_nasnet_large", "tf_pnasnet_large", "tf_resnet_v1_50", "tf_resnet_v1_101", "tf_resnet_v1_101",
-                                                "tf_resnet_v2_101", "tf_resnet_v2_152"};
-	
-    std::unordered_set<std::string> all_disabled_tests;
+    static const ORTCHAR_T* cuda_flaky_tests[] = {
+        ORT_TSTR("fp16_inception_v1"),
+        ORT_TSTR("fp16_shufflenet"), ORT_TSTR("fp16_tiny_yolov2")};
+    static const ORTCHAR_T* dml_disabled_tests[] = {ORT_TSTR("mlperf_ssd_resnet34_1200"), ORT_TSTR("mlperf_ssd_mobilenet_300"), ORT_TSTR("mask_rcnn_keras"), ORT_TSTR("mask_rcnn"), ORT_TSTR("faster_rcnn")};
+    static const ORTCHAR_T* dnnl_disabled_tests[] = {ORT_TSTR("test_densenet121"), ORT_TSTR("test_resnet18v2"), ORT_TSTR("test_resnet34v2"), ORT_TSTR("test_resnet50v2"), ORT_TSTR("test_resnet101v2"),
+                                                     ORT_TSTR("test_resnet101v2"), ORT_TSTR("test_vgg19"), ORT_TSTR("tf_inception_resnet_v2"), ORT_TSTR("tf_inception_v1"), ORT_TSTR("tf_inception_v3"), ORT_TSTR("tf_inception_v4"), ORT_TSTR("tf_mobilenet_v1_1.0_224"),
+                                                     ORT_TSTR("tf_mobilenet_v2_1.0_224"), ORT_TSTR("tf_mobilenet_v2_1.4_224"), ORT_TSTR("tf_nasnet_large"), ORT_TSTR("tf_pnasnet_large"), ORT_TSTR("tf_resnet_v1_50"), ORT_TSTR("tf_resnet_v1_101"), ORT_TSTR("tf_resnet_v1_101"),
+                                                     ORT_TSTR("tf_resnet_v2_101"), ORT_TSTR("tf_resnet_v2_152")};
+
+    std::unordered_set<std::basic_string<ORTCHAR_T> > all_disabled_tests;
     if (enable_cuda) {
       all_disabled_tests.insert(std::begin(cuda_flaky_tests), std::end(cuda_flaky_tests));
     }
@@ -373,25 +375,15 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
       // This will be removed after LRU implementation
       all_disabled_tests.insert(std::begin(dnnl_disabled_tests), std::end(dnnl_disabled_tests));
     }
-#if defined(__amd64__) || defined(_M_AMD64)
-#else
+#if !defined(__amd64__) && !defined(_M_AMD64)
     //out of memory
-    static const char* x86_disabled_tests[] = {"mlperf_ssd_resnet34_1200", "mask_rcnn_keras", "mask_rcnn", "faster_rcnn", "vgg19"};
+    static const ORTCHAR_T* x86_disabled_tests[] = {ORT_TSTR("mlperf_ssd_resnet34_1200"), ORT_TSTR("mask_rcnn_keras"), ORT_TSTR("mask_rcnn"), ORT_TSTR("faster_rcnn"), ORT_TSTR("vgg19")};
     all_disabled_tests.insert(std::begin(x86_disabled_tests), std::end(x86_disabled_tests));
 #endif
 
     std::vector<ITestCase*> tests;
-    LoadTests(data_dirs, whitelisted_test_cases, per_sample_tolerance, relative_per_sample_tolerance,
+    LoadTests(data_dirs, whitelisted_test_cases, per_sample_tolerance, relative_per_sample_tolerance, all_disabled_tests,
               [&tests](ITestCase* l) { tests.push_back(l); });
-    for (auto it = tests.begin(); it != tests.end();) {
-      auto iter = all_disabled_tests.find((*it)->GetTestCaseName());
-      if (iter != all_disabled_tests.end()) {
-        delete *it;
-        it = tests.erase(it);
-      } else {
-        ++it;
-      }
-    }
 
     TestEnv args(tests, stat, env, sf);
     Status st = RunTests(args, p_models, concurrent_session_runs, static_cast<size_t>(repeat_count),
diff --git a/onnxruntime/test/onnx/runner.cc b/onnxruntime/test/onnx/runner.cc
index d8942fc222138..3356f3b1b90fd 100644
--- a/onnxruntime/test/onnx/runner.cc
+++ b/onnxruntime/test/onnx/runner.cc
@@ -284,6 +284,7 @@ Status RunTests(TestEnv& env, int p_models, int concurrent_runs, size_t repeat_c
 void LoadTests(const std::vector<std::basic_string<PATH_CHAR_TYPE>>& input_paths,
                const std::vector<std::basic_string<PATH_CHAR_TYPE>>& whitelisted_test_cases,
                double default_per_sample_tolerance, double default_relative_per_sample_tolerance,
+               const std::unordered_set<std::basic_string<ORTCHAR_T>>& disabled_tests,
                const std::function<void(ITestCase*)>& process_function) {
   std::vector<std::basic_string<PATH_CHAR_TYPE>> paths(input_paths);
   while (!paths.empty()) {
@@ -302,10 +303,13 @@ void LoadTests(const std::vector<std::basic_string<PATH_CHAR_TYPE>>& input_paths
 
       std::basic_string<PATH_CHAR_TYPE> test_case_name = my_dir_name;
       if (test_case_name.compare(0, 5, ORT_TSTR("test_")) == 0) test_case_name = test_case_name.substr(5);
+
       if (!whitelisted_test_cases.empty() && std::find(whitelisted_test_cases.begin(), whitelisted_test_cases.end(),
                                                        test_case_name) == whitelisted_test_cases.end()) {
         return true;
       }
+      if (disabled_tests.find(test_case_name) != disabled_tests.end()) return true;
+
       std::basic_string<PATH_CHAR_TYPE> p = ConcatPathComponent<PATH_CHAR_TYPE>(node_data_root_path, filename_str);
 
       ITestCase* l = CreateOnnxTestCase(ToMBString(test_case_name), TestModelInfo::LoadOnnxModel(p.c_str()),
diff --git a/onnxruntime/test/onnx/runner.h b/onnxruntime/test/onnx/runner.h
index 9a26496b454f1..5f07b6bd78849 100644
--- a/onnxruntime/test/onnx/runner.h
+++ b/onnxruntime/test/onnx/runner.h
@@ -130,6 +130,7 @@ struct DataTask {
 void LoadTests(const std::vector<std::basic_string<PATH_CHAR_TYPE>>& input_paths,
                const std::vector<std::basic_string<PATH_CHAR_TYPE>>& whitelisted_test_cases,
                double default_per_sample_tolerance, double default_relative_per_sample_tolerance,
+               const std::unordered_set<std::basic_string<ORTCHAR_T>>& disabled_tests,
                const std::function<void(ITestCase*)>& process_function);
 
 //Do not run this function in the thread pool passed in
diff --git a/onnxruntime/test/util/compare_ortvalue.cc b/onnxruntime/test/util/compare_ortvalue.cc
index cfeaa11363185..958597997733c 100644
--- a/onnxruntime/test/util/compare_ortvalue.cc
+++ b/onnxruntime/test/util/compare_ortvalue.cc
@@ -5,11 +5,39 @@
 #include <cmath>
 #include <sstream>
 
+#ifdef __GNUC__
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wignored-qualifiers"
+#pragma GCC diagnostic ignored "-Wunused-parameter"
+#else
+#pragma warning(push)
+#pragma warning(disable : 4018) /*'expression' : signed/unsigned mismatch */
+#pragma warning(disable : 4065) /*switch statement contains 'default' but no 'case' labels*/
+#pragma warning(disable : 4100)
+#pragma warning(disable : 4146) /*unary minus operator applied to unsigned type, result still unsigned*/
+#pragma warning(disable : 4244) /*'conversion' conversion from 'type1' to 'type2', possible loss of data*/
+#pragma warning(disable : 4251) /*'identifier' : class 'type' needs to have dll-interface to be used by clients of class 'type2'*/
+#pragma warning(disable : 4267) /*'var' : conversion from 'size_t' to 'type', possible loss of data*/
+#pragma warning(disable : 4305) /*'identifier' : truncation from 'type1' to 'type2'*/
+#pragma warning(disable : 4307) /*'operator' : integral constant overflow*/
+#pragma warning(disable : 4309) /*'conversion' : truncation of constant value*/
+#pragma warning(disable : 4334) /*'operator' : result of 32-bit shift implicitly converted to 64 bits (was 64-bit shift intended?)*/
+#pragma warning(disable : 4355) /*'this' : used in base member initializer list*/
+#pragma warning(disable : 4506) /*no definition for inline function 'function'*/
+#pragma warning(disable : 4800) /*'type' : forcing value to bool 'true' or 'false' (performance warning)*/
+#pragma warning(disable : 4996) /*The compiler encountered a deprecated declaration.*/
+#endif
 #ifdef USE_FULL_PROTOBUF
 #include <google/protobuf/message.h>
 #else
 #include <google/protobuf/message_lite.h>
 #endif
+#ifdef __GNUC__
+#pragma GCC diagnostic pop
+#else
+#pragma warning(pop)
+#endif
+
 
 #include "core/graph/onnx_protobuf.h"
 #include "core/framework/tensorprotoutils.h"

From 56bb503c2f26474b6613bcb2a198691a11dcef38 Mon Sep 17 00:00:00 2001
From: Ashwini Khade <askhade@microsoft.com>
Date: Mon, 30 Dec 2019 12:31:44 -0800
Subject: [PATCH 16/75] Change default optimization level to All (from Basic)
 (#2745)

* change default optimization level to All (from Basic)

* fix test

* fix c# test
---
 csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs       | 4 ++--
 csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs | 2 +-
 csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs   | 2 +-
 docs/ONNX_Runtime_Graph_Optimizations.md                    | 4 +++-
 docs/ONNX_Runtime_Perf_Tuning.md                            | 4 ++--
 onnxruntime/core/framework/session_options.h                | 2 +-
 onnxruntime/python/onnxruntime_pybind_state.cc              | 6 +++---
 onnxruntime/test/onnx/main.cc                               | 2 +-
 onnxruntime/test/perftest/test_configuration.h              | 2 +-
 onnxruntime/test/python/onnxruntime_test_python.py          | 6 +++---
 10 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
index 70d48877da9c5..a21e16295b518 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
@@ -345,7 +345,7 @@ public int InterOpNumThreads
         private int _interOpNumThreads = 0; // set to what is set in C++ SessionOptions by default;
 
         /// <summary>
-        /// Sets the graph optimization level for the session. Default is set to ORT_ENABLE_BASIC.        
+        /// Sets the graph optimization level for the session. Default is set to ORT_ENABLE_ALL.
         /// </summary>
         public GraphOptimizationLevel GraphOptimizationLevel
         {
@@ -359,7 +359,7 @@ public GraphOptimizationLevel GraphOptimizationLevel
                 _graphOptimizationLevel = value;
             }
         }
-        private GraphOptimizationLevel _graphOptimizationLevel = GraphOptimizationLevel.ORT_ENABLE_BASIC;
+        private GraphOptimizationLevel _graphOptimizationLevel = GraphOptimizationLevel.ORT_ENABLE_ALL;
 
         /// <summary>
         /// Sets the execution mode for the session. Default is set to ORT_SEQUENTIAL.
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
index 29d3181568cd3..a18e4faf0846d 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
@@ -41,7 +41,7 @@ public void TestSessionOptions()
                 Assert.Equal(LogLevel.Verbose, opt.LogVerbosityLevel);
                 Assert.Equal(0, opt.IntraOpNumThreads);
                 Assert.Equal(0, opt.InterOpNumThreads);
-                Assert.Equal(GraphOptimizationLevel.ORT_ENABLE_BASIC, opt.GraphOptimizationLevel);
+                Assert.Equal(GraphOptimizationLevel.ORT_ENABLE_ALL, opt.GraphOptimizationLevel);
 
                 // try setting options 
                 opt.ExecutionMode = ExecutionMode.ORT_PARALLEL;
diff --git a/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs b/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
index de589114e7bad..e409f413e9e99 100644
--- a/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
+++ b/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
@@ -33,7 +33,7 @@ class CommandOptions
         public bool ParallelExecution { get; set; } = false;
 
         [Option('o', "optimization_level", Required = false, HelpText = "Optimization Level. Default is 1, partial optimization.")]
-        public GraphOptimizationLevel OptimizationLevel { get; set; } = GraphOptimizationLevel.ORT_ENABLE_BASIC;
+        public GraphOptimizationLevel OptimizationLevel { get; set; } = GraphOptimizationLevel.ORT_ENABLE_ALL;
     }
 
     class Program
diff --git a/docs/ONNX_Runtime_Graph_Optimizations.md b/docs/ONNX_Runtime_Graph_Optimizations.md
index 0c347f87e477d..2d124420db980 100644
--- a/docs/ONNX_Runtime_Graph_Optimizations.md
+++ b/docs/ONNX_Runtime_Graph_Optimizations.md
@@ -15,9 +15,11 @@ Graph optimizations are divided in three levels:
 
 The optimizations belonging to one level are performed after the optimizations of the previous level have been applied (e.g., extended optimizations are applied after basic optimizations have been applied).
 
+**All optimizations are enabled by default.**
+
 ### Basic Graph Optimizations
 
-These are semantics-preserving graph rewrites which remove redundant nodes and redundant computation. These optimizations are enabled by default. They run before graph partitioning and thus apply to all the execution providers. Available basic graph optimizations are as follows:
+These are semantics-preserving graph rewrites which remove redundant nodes and redundant computation. They run before graph partitioning and thus apply to all the execution providers. Available basic graph optimizations are as follows:
 
 * Constant Folding: Statically computes parts of the graph that rely only on constant initializers. This eliminates the need to compute them during runtime.
 
diff --git a/docs/ONNX_Runtime_Perf_Tuning.md b/docs/ONNX_Runtime_Perf_Tuning.md
index 2788ae041b85e..e2659700dec8f 100644
--- a/docs/ONNX_Runtime_Perf_Tuning.md
+++ b/docs/ONNX_Runtime_Perf_Tuning.md
@@ -86,9 +86,9 @@ sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL
 * Sequential vs Parallel Execution
   * `sess_options.execution_mode = rt.ExecutionMode.ORT_SEQUENTIAL` controls whether then operators in the graph should run sequentially or in parallel. Usually when a model has many branches, setting this option to false will provide better performance.
   * When `sess_options.execution_mode = rt.ExecutionMode.ORT_PARALLEL`, you can set `sess_options.inter_op_num_threads` to control the
-
 number of threads used to parallelize the execution of the graph (across nodes).
-* sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL. Default is ORT_ENABLE_BASIC(1). Please see [onnxruntime_c_api.h](../include/onnxruntime/core/session/onnxruntime_c_api.h#L241)  (enum GraphOptimizationLevel) for the full list of all optimization levels. For details regarding available optimizations and usage please refer to the [Graph Optimizations Doc](../docs/ONNX_Runtime_Graph_Optimizations.md).
+
+* sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL. Default is ORT_ENABLE_All. Please see [onnxruntime_c_api.h](../include/onnxruntime/core/session/onnxruntime_c_api.h#L241)  (enum GraphOptimizationLevel) for the full list of all optimization levels. For details regarding available optimizations and usage please refer to the [Graph Optimizations Doc](../docs/ONNX_Runtime_Graph_Optimizations.md).
 
 ### MKL_DNN/nGraph/MKL_ML Execution Provider
 MKL_DNN, MKL_ML and nGraph all depends on openmp for parallization. For those execution providers, we need to use the openmp enviroment variable to tune the performance.
diff --git a/onnxruntime/core/framework/session_options.h b/onnxruntime/core/framework/session_options.h
index aaefc50cc6c9c..1c02f9f0ebbc2 100644
--- a/onnxruntime/core/framework/session_options.h
+++ b/onnxruntime/core/framework/session_options.h
@@ -52,7 +52,7 @@ struct SessionOptions {
   unsigned max_num_graph_transformation_steps = 10;  // TODO choose a good default here?
 
   // set graph optimization level
-  TransformerLevel graph_optimization_level = TransformerLevel::Level1;
+  TransformerLevel graph_optimization_level = TransformerLevel::Level3;
 
   // controls the size of the thread pool used to parallelize the execution of tasks within individual nodes (ops)
   int intra_op_num_threads = 0;
diff --git a/onnxruntime/python/onnxruntime_pybind_state.cc b/onnxruntime/python/onnxruntime_pybind_state.cc
index e48c03323cc55..ba6035ba2da70 100644
--- a/onnxruntime/python/onnxruntime_pybind_state.cc
+++ b/onnxruntime/python/onnxruntime_pybind_state.cc
@@ -587,7 +587,7 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
       .def_property(
           "graph_optimization_level",
           [](const SessionOptions* options) -> GraphOptimizationLevel {
-            GraphOptimizationLevel retval = ORT_ENABLE_BASIC;
+            GraphOptimizationLevel retval = ORT_ENABLE_ALL;
             switch (options->graph_optimization_level) {
               case onnxruntime::TransformerLevel::Default:
                 retval = ORT_DISABLE_ALL;
@@ -602,8 +602,8 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
                 retval = ORT_ENABLE_ALL;
                 break;
               default:
-                retval = ORT_ENABLE_BASIC;
-                LOGS_DEFAULT(WARNING) << "Got invalid graph optimization level; defaulting to ORT_ENABLE_BASIC";
+                retval = ORT_ENABLE_ALL;
+                LOGS_DEFAULT(WARNING) << "Got invalid graph optimization level; defaulting to ORT_ENABLE_ALL";
                 break;
             }
             return retval;
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index 6fa0b30456d03..915212d818b33 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -41,7 +41,7 @@ void usage() {
       "Default: 'cpu'.\n"
       "\t-x: Use parallel executor, default (without -x): sequential executor.\n"
       "\t-d [device_id]: Specifies the device id for multi-device (e.g. GPU). The value should > 0\n"
-      "\t-o [optimization level]: Default is 1. Valid values are 0 (disable), 1 (basic), 2 (extended), 99 (all).\n"
+      "\t-o [optimization level]: Default is 99. Valid values are 0 (disable), 1 (basic), 2 (extended), 99 (all).\n"
       "\t\tPlease see onnxruntime_c_api.h (enum GraphOptimizationLevel) for the full list of all optimization levels. "
       "\n"
       "\t-h: help\n"
diff --git a/onnxruntime/test/perftest/test_configuration.h b/onnxruntime/test/perftest/test_configuration.h
index 9d0bf46774057..36186e129f0f8 100644
--- a/onnxruntime/test/perftest/test_configuration.h
+++ b/onnxruntime/test/perftest/test_configuration.h
@@ -47,7 +47,7 @@ struct RunConfig {
   ExecutionMode execution_mode{ExecutionMode::ORT_SEQUENTIAL};
   int intra_op_num_threads{0};
   int inter_op_num_threads{0};
-  GraphOptimizationLevel optimization_level{ORT_ENABLE_EXTENDED};
+  GraphOptimizationLevel optimization_level{ORT_ENABLE_ALL};
   std::basic_string<ORTCHAR_T> optimized_model_path;
 };
 
diff --git a/onnxruntime/test/python/onnxruntime_test_python.py b/onnxruntime/test/python/onnxruntime_test_python.py
index 0a0060767ebea..b8d5263f0301e 100644
--- a/onnxruntime/test/python/onnxruntime_test_python.py
+++ b/onnxruntime/test/python/onnxruntime_test_python.py
@@ -521,10 +521,10 @@ def test_run_model_mlnet(self):
 
     def testGraphOptimizationLevel(self):
         opt = onnxrt.SessionOptions()
-        self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_BASIC)
-            # default should be basic optimization
-        opt.graph_optimization_level = onnxrt.GraphOptimizationLevel.ORT_ENABLE_ALL
         self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_ALL)
+            # default should be all optimization
+        opt.graph_optimization_level = onnxrt.GraphOptimizationLevel.ORT_ENABLE_EXTENDED
+        self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_EXTENDED)
         sess = onnxrt.InferenceSession(self.get_name("logicaland.onnx"), sess_options=opt)
         a = np.array([[True, True], [False, False]], dtype=np.bool)
         b = np.array([[True, False], [True, False]], dtype=np.bool)

From fd334aff44b5e5dfe698716a31808c9d002586b3 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Mon, 30 Dec 2019 14:51:01 -0800
Subject: [PATCH 17/75] Update numpy to 1.18 (#2758)

* Update numpy to 1.18
---
 setup.py                                      | 11 +++---
 tools/ci_build/build.py                       |  2 +-
 .../azure-pipelines-py-packaging.yml          |  4 +-
 .../mac-nocontribops-ci-pipeline.yml          | 39 +++----------------
 .../linux/docker/scripts/requirements.txt     |  2 +-
 5 files changed, 15 insertions(+), 43 deletions(-)

diff --git a/setup.py b/setup.py
index 1347ccf318a91..c11c03f295bff 100644
--- a/setup.py
+++ b/setup.py
@@ -212,19 +212,20 @@ def run(self):
         'onnxruntime': data + examples + extra,
     },
     py_modules=python_modules_list,
-    extras_require={
-        'backend': ['onnx>=1.2.3'],
-        'numpy': ['numpy>=1.15.0']
-    },
+    install_requires=[
+        'onnx>=1.2.3',
+        'numpy>=1.18.0,<2'
+    ],
     entry_points= {
         'console_scripts': [
             'onnxruntime_test = onnxruntime.tools.onnxruntime_test:main',
         ]
     },
     classifiers=[
-        'Development Status :: 4 - Beta',
+        'Development Status :: 5 - Production/Stable',
         'Environment :: Console',
         'Intended Audience :: Developers',
+        'License :: OSI Approved :: MIT License',
         'Operating System :: POSIX :: Linux',
         'Programming Language :: Python',
         'Programming Language :: Python :: 3 :: Only',
diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index f420cb13cda81..09995aa63de8a 100755
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -244,7 +244,7 @@ def install_ubuntu_deps(args):
 
 def install_python_deps(numpy_version=""):
     dep_packages = ['setuptools', 'wheel', 'pytest']
-    dep_packages.append('numpy=={}'.format(numpy_version) if numpy_version else 'numpy>=1.15.0')
+    dep_packages.append('numpy=={}'.format(numpy_version) if numpy_version else 'numpy>=1.18.0')
     dep_packages.append('sympy>=1.1')
     dep_packages.append('packaging')
     run_subprocess([sys.executable, '-m', 'pip', 'install', '--trusted-host', 'files.pythonhosted.org'] + dep_packages)
diff --git a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
index 4054842b70eea..742784dd8c6b5 100644
--- a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
+++ b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
@@ -158,7 +158,7 @@ jobs:
       workingFolder: '$(Build.BinariesDirectory)'
 
   - script: |
-     python -m pip install -q pyopenssl setuptools wheel numpy     
+     python -m pip install -q pyopenssl setuptools wheel numpy==1.18
    
     workingDirectory: '$(Build.BinariesDirectory)'
     displayName: 'Install python modules'        
@@ -227,7 +227,7 @@ jobs:
 
 
     - script: |
-       python -m pip install -q pyopenssl setuptools wheel numpy     
+       python -m pip install -q pyopenssl setuptools wheel numpy==1.18
       workingDirectory: '$(Build.BinariesDirectory)'
       displayName: 'Install python modules'
 
diff --git a/tools/ci_build/github/azure-pipelines/mac-nocontribops-ci-pipeline.yml b/tools/ci_build/github/azure-pipelines/mac-nocontribops-ci-pipeline.yml
index 39aca7ec08596..5455b55a1639e 100644
--- a/tools/ci_build/github/azure-pipelines/mac-nocontribops-ci-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/mac-nocontribops-ci-pipeline.yml
@@ -1,35 +1,6 @@
-variables:
-  DisableContribOps: ON
-  TestDataUrlNoContribOps : https://onnxruntimetestdata.blob.core.windows.net/models/20181210.zip
-
 jobs:
-- job: MacOS_CI_Dev
-  pool:
-    vmImage: 'macOS-10.13'
-  steps:
-    - template: templates/set-test-data-variables-step.yml
-    - task: CmdLine@2
-      displayName: 'Download azcopy'
-      inputs:
-        script: |
-          curl -so azcopy.tar.gz -L 'https://aka.ms/downloadazcopy-v10-mac'
-          tar -zxvf azcopy.tar.gz --strip 1
-        workingDirectory: $(Build.BinariesDirectory)
-    - task: PythonScript@0
-      displayName: 'Download test data'
-      inputs:
-        scriptPath: '$(Build.SourcesDirectory)/tools/ci_build/github/download_test_data.py'
-        arguments: --test_data_url $(TestDataUrlNoContribOps) --azure_region centralus --build_dir $(Build.BinariesDirectory)
-        pythonInterpreter: '/usr/local/bin/python3'
-        workingDirectory: $(Build.BinariesDirectory)
-    - script: |
-        sudo python3 -m pip install numpy==1.15.0
-        sudo xcode-select --switch /Applications/Xcode_10.app/Contents/Developer
-        python3 $(Build.SourcesDirectory)/tools/ci_build/build.py --build_dir $(Build.BinariesDirectory) --enable_onnx_tests --skip_submodule_sync --parallel --build_shared_lib --disable_contrib_ops --config Debug Release
-      displayName: 'Build and Test OnnxRuntime lib for MacOS'
-
-    - task: ms.vss-governance-buildtask.governance-build-task-component-detection.ComponentGovernanceComponentDetection@0
-      displayName: 'Component Detection'
-      condition: and(succeeded(), in(variables['Build.Reason'], 'IndividualCI', 'BatchedCI'))
-
-    - template: templates/clean-agent-build-directory-step.yml
+- template: templates/mac-ci.yml
+  parameters:
+    AgentPool : 'Hosted macOS High Sierra'
+    DoNugetPack: 'false'
+    BuildCommand: 'python3 $(Build.SourcesDirectory)/tools/ci_build/build.py --use_openmp --build_dir $(Build.BinariesDirectory) --build_wheel --skip_submodule_sync --parallel --build_shared_lib --disable_contrib_ops --enable_onnx_tests --config Debug RelWithDebInfo'
diff --git a/tools/ci_build/github/linux/docker/scripts/requirements.txt b/tools/ci_build/github/linux/docker/scripts/requirements.txt
index f41acbfb8e766..40c0bcb0da95b 100644
--- a/tools/ci_build/github/linux/docker/scripts/requirements.txt
+++ b/tools/ci_build/github/linux/docker/scripts/requirements.txt
@@ -1,4 +1,4 @@
-numpy==1.15.0
+numpy==1.18.0
 requests==2.21.0
 scipy
 mypy

From 382fa86af8703db4f5868d511632592d2058f568 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Thu, 2 Jan 2020 15:25:25 -0800
Subject: [PATCH 18/75] Pipeline changes for python 3.8 (#2753)

1. Pipeline changes for python 3.8
2. Fix a regression in setup.py which was just introduced in the previous commit.

Please notice, we still haven't made python 3.8 + Windows + CUDA work.
---
 setup.py                                      |  2 +-
 .../azure-pipelines-py-packaging.yml          | 24 ++++++++++++++++---
 .../linux/docker/scripts/install_deps.sh      |  2 ++
 .../linux/docker/scripts/install_onnx.sh      |  8 +++----
 4 files changed, 28 insertions(+), 8 deletions(-)

diff --git a/setup.py b/setup.py
index c11c03f295bff..4f3ba6e535e41 100644
--- a/setup.py
+++ b/setup.py
@@ -214,7 +214,7 @@ def run(self):
     py_modules=python_modules_list,
     install_requires=[
         'onnx>=1.2.3',
-        'numpy>=1.18.0,<2'
+        'numpy>=1.18.0'
     ],
     entry_points= {
         'console_scripts': [
diff --git a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
index 742784dd8c6b5..c2bafb57fe6be 100644
--- a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
+++ b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
@@ -8,12 +8,19 @@ jobs:
       Python35:
         python.version: '3.5'
         python.dir: '/opt/python/cp35-cp35m'
+        python.include.dir: '/opt/python/cp35-cp35m/include/python3.5m'
       Python36:
         python.version: '3.6'
         python.dir: '/opt/python/cp36-cp36m'
+        python.include.dir: '/opt/python/cp36-cp36m/include/python3.6m'
       Python37:
         python.version: '3.7'
         python.dir: '/opt/python/cp37-cp37m'
+        python.include.dir: '/opt/python/cp37-cp37m/include/python3.7m'
+      Python38:
+        python.version: '3.8'
+        python.dir: '/opt/python/cp38-cp38'
+        python.include.dir: '/opt/python/cp38-cp38/include/python3.8'
   steps:
     - template: templates/set-test-data-variables-step.yml
 
@@ -40,7 +47,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-manylinux-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.dir)/include/python$(python.version)m PYTHON_LIBRARY=/usr/lib64/librt.so
+          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-manylinux-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so
         workingDirectory: $(Build.SourcesDirectory)
 
     - task: CopyFiles@2
@@ -69,12 +76,19 @@ jobs:
       Python35:
         python.version: '3.5'
         python.dir: '/opt/python/cp35-cp35m'
+        python.include.dir: '/opt/python/cp35-cp35m/include/python3.5m'
       Python36:
         python.version: '3.6'
         python.dir: '/opt/python/cp36-cp36m'
+        python.include.dir: '/opt/python/cp36-cp36m/include/python3.6m'
       Python37:
         python.version: '3.7'
         python.dir: '/opt/python/cp37-cp37m'
+        python.include.dir: '/opt/python/cp37-cp37m/include/python3.7m'
+      Python38:
+        python.version: '3.8'
+        python.dir: '/opt/python/cp38-cp38'
+        python.include.dir: '/opt/python/cp38-cp38/include/python3.8'
   steps:
     - template: templates/set-test-data-variables-step.yml
 
@@ -101,7 +115,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-manylinux-gpu-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.dir)/include/python$(python.version)m PYTHON_LIBRARY=/usr/lib64/librt.so --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0
+          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-manylinux-gpu-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0
         workingDirectory: $(Build.SourcesDirectory)
 
     - task: CopyFiles@2
@@ -131,6 +145,8 @@ jobs:
         python.version: '3.6'
       Python37:
         python.version: '3.7'
+      Python38:
+        python.version: '3.8'
   variables:
     OrtPackageId: 'Microsoft.ML.OnnxRuntime'
     MsbuildArguments: '-maxcpucount'
@@ -199,7 +215,7 @@ jobs:
   workspace:
     clean: all
   pool: 'Win-GPU-2019'
-  timeoutInMinutes:  120  
+  timeoutInMinutes:  60  
   variables:
     CUDA_VERSION: '10.0'
     EnvSetupScript: setup_env.bat
@@ -276,6 +292,8 @@ jobs:
         python.version: '3.6'
       Python37:
         python.version: '3.7'
+      Python38:
+        python.version: '3.8'
   steps:
     - task: UsePythonVersion@0
       displayName: 'Use Python'
diff --git a/tools/ci_build/github/linux/docker/scripts/install_deps.sh b/tools/ci_build/github/linux/docker/scripts/install_deps.sh
index 768c4f7852069..bf39742be9785 100755
--- a/tools/ci_build/github/linux/docker/scripts/install_deps.sh
+++ b/tools/ci_build/github/linux/docker/scripts/install_deps.sh
@@ -54,6 +54,8 @@ elif [[ "$PYTHON_VER" = "3.6" && -d "/opt/python/cp36-cp36m"  ]]; then
    PYTHON_EXE="/opt/python/cp36-cp36m/bin/python3.6"
 elif [[ "$PYTHON_VER" = "3.7" && -d "/opt/python/cp37-cp37m"  ]]; then
    PYTHON_EXE="/opt/python/cp37-cp37m/bin/python3.7"
+elif [[ "$PYTHON_VER" = "3.8" && -d "/opt/python/cp38-cp38"  ]]; then
+   PYTHON_EXE="/opt/python/cp38-cp38/bin/python3.8"
 else
    PYTHON_EXE="/usr/bin/python${PYTHON_VER}"
 fi
diff --git a/tools/ci_build/github/linux/docker/scripts/install_onnx.sh b/tools/ci_build/github/linux/docker/scripts/install_onnx.sh
index fc2cc11907c97..20b637f159082 100755
--- a/tools/ci_build/github/linux/docker/scripts/install_onnx.sh
+++ b/tools/ci_build/github/linux/docker/scripts/install_onnx.sh
@@ -14,13 +14,12 @@ PYTHON_VER=$1
 
 if [[ "$PYTHON_VER" = "3.5" && -d "/opt/python/cp35-cp35m"  ]]; then
    PYTHON_EXE="/opt/python/cp35-cp35m/bin/python3.5"
-   export PATH=/opt/python/cp35-cp35m/bin:$PATH
 elif [[ "$PYTHON_VER" = "3.6" && -d "/opt/python/cp36-cp36m"  ]]; then
    PYTHON_EXE="/opt/python/cp36-cp36m/bin/python3.6"
-   export PATH=/opt/python/cp36-cp36m/bin:$PATH
 elif [[ "$PYTHON_VER" = "3.7" && -d "/opt/python/cp37-cp37m"  ]]; then
    PYTHON_EXE="/opt/python/cp37-cp37m/bin/python3.7"
-   export PATH=/opt/python/cp37-cp37m/bin:$PATH
+elif [[ "$PYTHON_VER" = "3.8" && -d "/opt/python/cp38-cp38"  ]]; then
+   PYTHON_EXE="/opt/python/cp38-cp38/bin/python3.8"
 else
    PYTHON_EXE="/usr/bin/python${PYTHON_VER}"
 fi
@@ -57,5 +56,6 @@ for v2t in ${version2tag[*]}; do
   mv /tmp/src/onnx-$onnx_version/tools/repl_protoc-gen-mypy.py /tmp/src/onnx-$onnx_version/tools/protoc-gen-mypy.py
   mkdir -p /data/onnx/${onnx_tag}
   ${PYTHON_EXE} -m pip install .
-  backend-test-tools generate-data -o /data/onnx/$onnx_tag
+  cd /tmp  
+  ${PYTHON_EXE} -m onnx.backend.test.cmd_tools generate-data -o /data/onnx/$onnx_tag
 done

From a2c8981a9e01aa87d2ee38a113ad6bbffd218bd3 Mon Sep 17 00:00:00 2001
From: Scott McKay <skottmckay@gmail.com>
Date: Fri, 3 Jan 2020 13:43:57 +1000
Subject: [PATCH 19/75] Add basic stacktrace output for posix debug builds.
 (#2749)

---
 onnxruntime/core/platform/posix/stacktrace.cc | 38 ++++++++++++++++++-
 1 file changed, 36 insertions(+), 2 deletions(-)

diff --git a/onnxruntime/core/platform/posix/stacktrace.cc b/onnxruntime/core/platform/posix/stacktrace.cc
index f41ef0e1d4902..73a4e2b274f62 100644
--- a/onnxruntime/core/platform/posix/stacktrace.cc
+++ b/onnxruntime/core/platform/posix/stacktrace.cc
@@ -3,10 +3,44 @@
 
 #include "core/common/common.h"
 
+#include <execinfo.h>
+#include <vector>
+
 namespace onnxruntime {
 
 std::vector<std::string> GetStackTrace() {
-  return {"<stacktrace not implemented>"};
-}
+  std::vector<std::string> stack;
+
+#ifndef NDEBUG
+  constexpr int kCallstackLimit = 64;  // Maximum depth of callstack
+
+  void* array[kCallstackLimit];
+  char** strings = nullptr;
+
+  size_t size = backtrace(array, kCallstackLimit);
+  stack.reserve(size);
+  strings = backtrace_symbols(array, size);
 
+  // NOTE: To get meaningful info from the output, addr2line (or atos on osx) would need to be used.
+  // See https://gist.github.com/jvranish/4441299 for an example.
+  //
+  // To manually translate the output, use the value in the '()' after the executable name with addr2line
+  // e.g.
+  //   Stacktrace:
+  //    /home/me/src/github/onnxruntime/build/Linux/Debug/onnxruntime_test_all(+0x3f46cc) [0x559543faf6cc]
+  //
+  // >addr2line -f -C -e /home/me/src/github/onnxruntime/build/Linux/Debug/onnxruntime_test_all  +0x3f46cc
+
+  // hide GetStackTrace so the output starts with the 'real' location
+  constexpr size_t start_frame = 1;
+  for (size_t i = start_frame; i < size; i++) {
+    stack.push_back(strings[i]);
+  }
+
+  free(strings);
+
+#endif
+
+  return stack;
+}
 }  // namespace onnxruntime

From 2096f324f436a8ba7861f9434935f68e37e104d6 Mon Sep 17 00:00:00 2001
From: KeDengMS <kedeng@microsoft.com>
Date: Fri, 3 Jan 2020 10:05:09 -0800
Subject: [PATCH 20/75] [NupharEP] fix a race condition when multiple sessions
 running different models concurrently (#2772)

---
 .../core/providers/nuphar/nuphar_execution_provider.cc    | 8 ++++----
 .../core/providers/nuphar/nuphar_execution_provider.h     | 7 ++++---
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/onnxruntime/core/providers/nuphar/nuphar_execution_provider.cc b/onnxruntime/core/providers/nuphar/nuphar_execution_provider.cc
index 6e76bc45d0ad9..1f61783b775f5 100644
--- a/onnxruntime/core/providers/nuphar/nuphar_execution_provider.cc
+++ b/onnxruntime/core/providers/nuphar/nuphar_execution_provider.cc
@@ -25,7 +25,7 @@ namespace onnxruntime {
 thread_local int64_t NupharSubgraphUnit::counter = 0;
 
 thread_local std::unique_ptr<std::unordered_map<std::string, int64_t>> NupharExecutionProvider::tls_realized_dims_;
-int NupharExecutionProvider::global_fused_count_ = 0;
+thread_local int NupharExecutionProvider::per_model_fused_count_ = 0;
 
 static std::string GetCurrentHostTargetString() {
 #if USE_TVM_WITH_LLVM
@@ -312,12 +312,12 @@ NupharExecutionProvider::GetCapability(const onnxruntime::GraphViewer& graph_vie
   };
   GraphPartitioner graph_partitioner(is_supported_func);
 
-  ORT_ENFORCE(graph_partitioner.Partition(graph_viewer, global_fused_count_, results).IsOK());
+  ORT_ENFORCE(graph_partitioner.Partition(graph_viewer, per_model_fused_count_, results).IsOK());
 
-  // reset global_fused_count_ for main graph, since there might be multiple sessions for subgraphs,
+  // reset per_model_fused_count_ for main graph, since there might be multiple sessions for subgraphs,
   // this is the time all graph cut should be finished as ORT handles main graph last
   if (!graph_viewer.IsSubgraph())
-    global_fused_count_ = 0;
+    per_model_fused_count_ = 0;
 
   // for any node being fused in results, save initializer tensors
   // because IExecutionProvider::Compile would be called without OpKernelInfo
diff --git a/onnxruntime/core/providers/nuphar/nuphar_execution_provider.h b/onnxruntime/core/providers/nuphar/nuphar_execution_provider.h
index edfb5aa14ffc3..67c26194f1498 100644
--- a/onnxruntime/core/providers/nuphar/nuphar_execution_provider.h
+++ b/onnxruntime/core/providers/nuphar/nuphar_execution_provider.h
@@ -154,9 +154,10 @@ class NupharExecutionProvider : public IExecutionProvider {
   mutable std::unordered_map<std::string, std::unique_ptr<Tensor>> constant_initializers_used_in_compiled_nodes_;
   mutable std::unordered_map<std::string, int> domain_versions_;
 
-  // used to create unique fused node name, make it static because
-  // subsession may create multiple instances of EPs
-  static int global_fused_count_;
+  // used to create unique fused node name, make it thread_local because
+  // subsession of a model with subgraph may create multiple instances of EPs,
+  // and there might be multiple inference sessions running different models concurrently
+  static thread_local int per_model_fused_count_;
 };
 
 }  // namespace onnxruntime

From 013642ed37b1bf78dbc17bbc9d8893a906914279 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 3 Jan 2020 13:35:32 -0800
Subject: [PATCH 21/75] Revert "Change default optimization level to All (from
 Basic) (#2745)"

This reverts commit 56bb503c2f26474b6613bcb2a198691a11dcef38.
---
 csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs       | 4 ++--
 csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs | 2 +-
 csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs   | 2 +-
 docs/ONNX_Runtime_Graph_Optimizations.md                    | 4 +---
 docs/ONNX_Runtime_Perf_Tuning.md                            | 4 ++--
 onnxruntime/core/framework/session_options.h                | 2 +-
 onnxruntime/python/onnxruntime_pybind_state.cc              | 6 +++---
 onnxruntime/test/onnx/main.cc                               | 2 +-
 onnxruntime/test/perftest/test_configuration.h              | 2 +-
 onnxruntime/test/python/onnxruntime_test_python.py          | 6 +++---
 10 files changed, 16 insertions(+), 18 deletions(-)

diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
index a21e16295b518..70d48877da9c5 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
@@ -345,7 +345,7 @@ public int InterOpNumThreads
         private int _interOpNumThreads = 0; // set to what is set in C++ SessionOptions by default;
 
         /// <summary>
-        /// Sets the graph optimization level for the session. Default is set to ORT_ENABLE_ALL.
+        /// Sets the graph optimization level for the session. Default is set to ORT_ENABLE_BASIC.        
         /// </summary>
         public GraphOptimizationLevel GraphOptimizationLevel
         {
@@ -359,7 +359,7 @@ public GraphOptimizationLevel GraphOptimizationLevel
                 _graphOptimizationLevel = value;
             }
         }
-        private GraphOptimizationLevel _graphOptimizationLevel = GraphOptimizationLevel.ORT_ENABLE_ALL;
+        private GraphOptimizationLevel _graphOptimizationLevel = GraphOptimizationLevel.ORT_ENABLE_BASIC;
 
         /// <summary>
         /// Sets the execution mode for the session. Default is set to ORT_SEQUENTIAL.
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
index a18e4faf0846d..29d3181568cd3 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
@@ -41,7 +41,7 @@ public void TestSessionOptions()
                 Assert.Equal(LogLevel.Verbose, opt.LogVerbosityLevel);
                 Assert.Equal(0, opt.IntraOpNumThreads);
                 Assert.Equal(0, opt.InterOpNumThreads);
-                Assert.Equal(GraphOptimizationLevel.ORT_ENABLE_ALL, opt.GraphOptimizationLevel);
+                Assert.Equal(GraphOptimizationLevel.ORT_ENABLE_BASIC, opt.GraphOptimizationLevel);
 
                 // try setting options 
                 opt.ExecutionMode = ExecutionMode.ORT_PARALLEL;
diff --git a/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs b/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
index e409f413e9e99..de589114e7bad 100644
--- a/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
+++ b/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
@@ -33,7 +33,7 @@ class CommandOptions
         public bool ParallelExecution { get; set; } = false;
 
         [Option('o', "optimization_level", Required = false, HelpText = "Optimization Level. Default is 1, partial optimization.")]
-        public GraphOptimizationLevel OptimizationLevel { get; set; } = GraphOptimizationLevel.ORT_ENABLE_ALL;
+        public GraphOptimizationLevel OptimizationLevel { get; set; } = GraphOptimizationLevel.ORT_ENABLE_BASIC;
     }
 
     class Program
diff --git a/docs/ONNX_Runtime_Graph_Optimizations.md b/docs/ONNX_Runtime_Graph_Optimizations.md
index 2d124420db980..0c347f87e477d 100644
--- a/docs/ONNX_Runtime_Graph_Optimizations.md
+++ b/docs/ONNX_Runtime_Graph_Optimizations.md
@@ -15,11 +15,9 @@ Graph optimizations are divided in three levels:
 
 The optimizations belonging to one level are performed after the optimizations of the previous level have been applied (e.g., extended optimizations are applied after basic optimizations have been applied).
 
-**All optimizations are enabled by default.**
-
 ### Basic Graph Optimizations
 
-These are semantics-preserving graph rewrites which remove redundant nodes and redundant computation. They run before graph partitioning and thus apply to all the execution providers. Available basic graph optimizations are as follows:
+These are semantics-preserving graph rewrites which remove redundant nodes and redundant computation. These optimizations are enabled by default. They run before graph partitioning and thus apply to all the execution providers. Available basic graph optimizations are as follows:
 
 * Constant Folding: Statically computes parts of the graph that rely only on constant initializers. This eliminates the need to compute them during runtime.
 
diff --git a/docs/ONNX_Runtime_Perf_Tuning.md b/docs/ONNX_Runtime_Perf_Tuning.md
index e2659700dec8f..2788ae041b85e 100644
--- a/docs/ONNX_Runtime_Perf_Tuning.md
+++ b/docs/ONNX_Runtime_Perf_Tuning.md
@@ -86,9 +86,9 @@ sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL
 * Sequential vs Parallel Execution
   * `sess_options.execution_mode = rt.ExecutionMode.ORT_SEQUENTIAL` controls whether then operators in the graph should run sequentially or in parallel. Usually when a model has many branches, setting this option to false will provide better performance.
   * When `sess_options.execution_mode = rt.ExecutionMode.ORT_PARALLEL`, you can set `sess_options.inter_op_num_threads` to control the
-number of threads used to parallelize the execution of the graph (across nodes).
 
-* sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL. Default is ORT_ENABLE_All. Please see [onnxruntime_c_api.h](../include/onnxruntime/core/session/onnxruntime_c_api.h#L241)  (enum GraphOptimizationLevel) for the full list of all optimization levels. For details regarding available optimizations and usage please refer to the [Graph Optimizations Doc](../docs/ONNX_Runtime_Graph_Optimizations.md).
+number of threads used to parallelize the execution of the graph (across nodes).
+* sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL. Default is ORT_ENABLE_BASIC(1). Please see [onnxruntime_c_api.h](../include/onnxruntime/core/session/onnxruntime_c_api.h#L241)  (enum GraphOptimizationLevel) for the full list of all optimization levels. For details regarding available optimizations and usage please refer to the [Graph Optimizations Doc](../docs/ONNX_Runtime_Graph_Optimizations.md).
 
 ### MKL_DNN/nGraph/MKL_ML Execution Provider
 MKL_DNN, MKL_ML and nGraph all depends on openmp for parallization. For those execution providers, we need to use the openmp enviroment variable to tune the performance.
diff --git a/onnxruntime/core/framework/session_options.h b/onnxruntime/core/framework/session_options.h
index 1c02f9f0ebbc2..aaefc50cc6c9c 100644
--- a/onnxruntime/core/framework/session_options.h
+++ b/onnxruntime/core/framework/session_options.h
@@ -52,7 +52,7 @@ struct SessionOptions {
   unsigned max_num_graph_transformation_steps = 10;  // TODO choose a good default here?
 
   // set graph optimization level
-  TransformerLevel graph_optimization_level = TransformerLevel::Level3;
+  TransformerLevel graph_optimization_level = TransformerLevel::Level1;
 
   // controls the size of the thread pool used to parallelize the execution of tasks within individual nodes (ops)
   int intra_op_num_threads = 0;
diff --git a/onnxruntime/python/onnxruntime_pybind_state.cc b/onnxruntime/python/onnxruntime_pybind_state.cc
index ba6035ba2da70..e48c03323cc55 100644
--- a/onnxruntime/python/onnxruntime_pybind_state.cc
+++ b/onnxruntime/python/onnxruntime_pybind_state.cc
@@ -587,7 +587,7 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
       .def_property(
           "graph_optimization_level",
           [](const SessionOptions* options) -> GraphOptimizationLevel {
-            GraphOptimizationLevel retval = ORT_ENABLE_ALL;
+            GraphOptimizationLevel retval = ORT_ENABLE_BASIC;
             switch (options->graph_optimization_level) {
               case onnxruntime::TransformerLevel::Default:
                 retval = ORT_DISABLE_ALL;
@@ -602,8 +602,8 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
                 retval = ORT_ENABLE_ALL;
                 break;
               default:
-                retval = ORT_ENABLE_ALL;
-                LOGS_DEFAULT(WARNING) << "Got invalid graph optimization level; defaulting to ORT_ENABLE_ALL";
+                retval = ORT_ENABLE_BASIC;
+                LOGS_DEFAULT(WARNING) << "Got invalid graph optimization level; defaulting to ORT_ENABLE_BASIC";
                 break;
             }
             return retval;
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index 915212d818b33..6fa0b30456d03 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -41,7 +41,7 @@ void usage() {
       "Default: 'cpu'.\n"
       "\t-x: Use parallel executor, default (without -x): sequential executor.\n"
       "\t-d [device_id]: Specifies the device id for multi-device (e.g. GPU). The value should > 0\n"
-      "\t-o [optimization level]: Default is 99. Valid values are 0 (disable), 1 (basic), 2 (extended), 99 (all).\n"
+      "\t-o [optimization level]: Default is 1. Valid values are 0 (disable), 1 (basic), 2 (extended), 99 (all).\n"
       "\t\tPlease see onnxruntime_c_api.h (enum GraphOptimizationLevel) for the full list of all optimization levels. "
       "\n"
       "\t-h: help\n"
diff --git a/onnxruntime/test/perftest/test_configuration.h b/onnxruntime/test/perftest/test_configuration.h
index 36186e129f0f8..9d0bf46774057 100644
--- a/onnxruntime/test/perftest/test_configuration.h
+++ b/onnxruntime/test/perftest/test_configuration.h
@@ -47,7 +47,7 @@ struct RunConfig {
   ExecutionMode execution_mode{ExecutionMode::ORT_SEQUENTIAL};
   int intra_op_num_threads{0};
   int inter_op_num_threads{0};
-  GraphOptimizationLevel optimization_level{ORT_ENABLE_ALL};
+  GraphOptimizationLevel optimization_level{ORT_ENABLE_EXTENDED};
   std::basic_string<ORTCHAR_T> optimized_model_path;
 };
 
diff --git a/onnxruntime/test/python/onnxruntime_test_python.py b/onnxruntime/test/python/onnxruntime_test_python.py
index b8d5263f0301e..0a0060767ebea 100644
--- a/onnxruntime/test/python/onnxruntime_test_python.py
+++ b/onnxruntime/test/python/onnxruntime_test_python.py
@@ -521,10 +521,10 @@ def test_run_model_mlnet(self):
 
     def testGraphOptimizationLevel(self):
         opt = onnxrt.SessionOptions()
+        self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_BASIC)
+            # default should be basic optimization
+        opt.graph_optimization_level = onnxrt.GraphOptimizationLevel.ORT_ENABLE_ALL
         self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_ALL)
-            # default should be all optimization
-        opt.graph_optimization_level = onnxrt.GraphOptimizationLevel.ORT_ENABLE_EXTENDED
-        self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_EXTENDED)
         sess = onnxrt.InferenceSession(self.get_name("logicaland.onnx"), sess_options=opt)
         a = np.array([[True, True], [False, False]], dtype=np.bool)
         b = np.array([[True, False], [True, False]], dtype=np.bool)

From 3d75c1b6be50b552e2734298d1355668a4ed3cd9 Mon Sep 17 00:00:00 2001
From: Takeshi Watanabe <take-cheeze@users.noreply.github.com>
Date: Mon, 6 Jan 2020 06:59:23 +0900
Subject: [PATCH 22/75] Fix typo in error message (#2736)

---
 onnxruntime/core/providers/cpu/nn/tfidfvectorizer.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/onnxruntime/core/providers/cpu/nn/tfidfvectorizer.cc b/onnxruntime/core/providers/cpu/nn/tfidfvectorizer.cc
index 176a9d5f4356d..7e9a4636f1afc 100644
--- a/onnxruntime/core/providers/cpu/nn/tfidfvectorizer.cc
+++ b/onnxruntime/core/providers/cpu/nn/tfidfvectorizer.cc
@@ -388,7 +388,7 @@ TfIdfVectorizer::TfIdfVectorizer(const OpKernelInfo& info) : OpKernel(info), imp
         } else {
           auto before_insert = impl_->str_set_.size();
           Emplace(impl_->pool_strings_.begin() + start_idx, ngrams, ngram_size, ngram_id, impl_->str_set_);
-          ORT_ENFORCE((before_insert + ngrams) == impl_->str_set_.size(), "poll_strings duplicate ", std::to_string(ngram_size), "-grams detected");
+          ORT_ENFORCE((before_insert + ngrams) == impl_->str_set_.size(), "pool_strings duplicate ", std::to_string(ngram_size), "-grams detected");
         }
       } else {
         ngram_id += ngrams;

From 7666d130e51873d1ff92f86670ba04ebcbf9496b Mon Sep 17 00:00:00 2001
From: Changyoung Koh <gkcy1019@gmail.com>
Date: Tue, 7 Jan 2020 03:50:42 +0900
Subject: [PATCH 23/75] Rename MKL-DNN to DNNL to fix broken link (#2730)

---
 .../{MKL-DNN-ExecutionProvider.md => DNNL-ExecutionProvider.md}   | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename docs/execution_providers/{MKL-DNN-ExecutionProvider.md => DNNL-ExecutionProvider.md} (100%)

diff --git a/docs/execution_providers/MKL-DNN-ExecutionProvider.md b/docs/execution_providers/DNNL-ExecutionProvider.md
similarity index 100%
rename from docs/execution_providers/MKL-DNN-ExecutionProvider.md
rename to docs/execution_providers/DNNL-ExecutionProvider.md

From 7ace7a5bcd0b65b91b1bb3462118343663633a98 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Thu, 2 Jan 2020 16:45:31 -0800
Subject: [PATCH 24/75] Pass BUILD_BUILDNUMBER to linux docker

---
 .../github/azure-pipelines/azure-pipelines-py-packaging.yml   | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
index c2bafb57fe6be..e343448fce6b7 100644
--- a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
+++ b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
@@ -47,7 +47,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-manylinux-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so
+          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD -e BUILD_BUILDNUMBER onnxruntime-manylinux-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so
         workingDirectory: $(Build.SourcesDirectory)
 
     - task: CopyFiles@2
@@ -115,7 +115,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-manylinux-gpu-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0
+          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD -e BUILD_BUILDNUMBER onnxruntime-manylinux-gpu-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0
         workingDirectory: $(Build.SourcesDirectory)
 
     - task: CopyFiles@2

From e3f674b5635ca360b111ab30eecc87602c3cfb66 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 3 Jan 2020 12:00:03 -0800
Subject: [PATCH 25/75] Disable featurizers in python packages

---
 .../github/azure-pipelines/azure-pipelines-py-packaging.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
index e343448fce6b7..529306c6c5ec6 100644
--- a/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
+++ b/tools/ci_build/github/azure-pipelines/azure-pipelines-py-packaging.yml
@@ -47,7 +47,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD -e BUILD_BUILDNUMBER onnxruntime-manylinux-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so
+          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD -e BUILD_BUILDNUMBER onnxruntime-manylinux-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest  --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so
         workingDirectory: $(Build.SourcesDirectory)
 
     - task: CopyFiles@2
@@ -115,7 +115,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD -e BUILD_BUILDNUMBER onnxruntime-manylinux-gpu-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0
+          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD -e BUILD_BUILDNUMBER onnxruntime-manylinux-gpu-$(python.version) $(python.dir)/bin/python3 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest  --build_wheel --enable_onnx_tests --cmake_extra_defines PYTHON_INCLUDE_DIR=$(python.include.dir) PYTHON_LIBRARY=/usr/lib64/librt.so --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0
         workingDirectory: $(Build.SourcesDirectory)
 
     - task: CopyFiles@2
@@ -190,7 +190,7 @@ jobs:
     displayName: 'BUILD'
     inputs:
       scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
-      arguments: '--config RelWithDebInfo --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_generator "Visual Studio 16 2019" --build_wheel --use_featurizers --enable_onnx_tests --parallel'
+      arguments: '--config RelWithDebInfo --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_generator "Visual Studio 16 2019" --build_wheel  --enable_onnx_tests --parallel'
       workingDirectory: '$(Build.BinariesDirectory)'     
   
   - task: CopyFiles@2

From 1b231180561cdfa9d0e2b74a511e050018b4ba58 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Thu, 2 Jan 2020 15:43:40 -0800
Subject: [PATCH 26/75] Fix nightly build version number issue

---
 setup.py | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index 4f3ba6e535e41..e0dfb4e729883 100644
--- a/setup.py
+++ b/setup.py
@@ -183,7 +183,13 @@ def run(self):
 with open('VERSION_NUMBER') as f:
     version_number = f.readline().strip()
 if nightly_build:
-    date_suffix = str(datetime.datetime.now().date().strftime("%m%d"))
+    #https://docs.microsoft.com/en-us/azure/devops/pipelines/build/variables
+    date_suffix = environ.get('BUILD_BUILDNUMBER')
+    if date_suffix is None:
+      #The following line is only for local testing
+      date_suffix = str(datetime.datetime.now().date().strftime("%Y%m%d"))
+    else:
+      date_suffix = date_suffix.replace('.','')
     version_number = version_number + ".dev" + date_suffix
 
 cmd_classes = {}

From 6f66260372bfa0e6f804eb7fe95c2d275a1442ec Mon Sep 17 00:00:00 2001
From: Dmitri Smirnov <yuslepukhin@users.noreply.github.com>
Date: Mon, 6 Jan 2020 13:43:44 -0800
Subject: [PATCH 27/75] Import more featurizers (#2781)

Make kernels non-template. Add input constraint for learnt data.
  Add min_max_scalar_transformer, robust_scalar_transformer,
  inputation_marker_transfomer, label_encoder_transformer,
 missing_dummies_transformer along with tests.
 Advance Featurizers library commit.
---
 cgmanifest.json                               |   2 +-
 cmake/external/featurizers.cmake              |   9 +-
 .../graph/featurizers_ops/featurizers_defs.cc | 399 ++++++++++++++++--
 .../cpu/cat_imputer_transformer.cc            |  51 +--
 .../cpu/date_time_transformer.cc              |   2 +-
 .../cpu/imputation_marker_transformer.cc      |  76 ++++
 .../cpu/label_encoder_transformer.cc          |  80 ++++
 .../cpu/max_abs_scalar_transformer.cc         | 128 ++----
 .../cpu/min_max_scalar_transformer.cc         |  78 ++++
 .../cpu/missing_dummies_transformer.cc        |  77 ++++
 .../cpu/robust_scalar_transformer.cc          | 102 +++++
 .../featurizers_ops/cpu/string_transformer.cc | 150 ++-----
 .../cpu_featurizers_kernels.cc                |  69 +--
 .../featurizers_ops/categoryimputer_test.cc   |  14 +-
 .../datetimetransformer_test.cc               |   8 +-
 .../imputationmarkertransformer_test.cc       | 146 +++++++
 .../labelencodertransfomer_test.cc            | 110 +++++
 ...est.cc => maxabsscalertransformer_test.cc} |  10 +-
 .../minmaxscalartransformer_test.cc           |  60 +++
 .../missingdummiestransfomer_test.cc          |  58 +++
 .../robustscalartransformer_test.cc           | 109 +++++
 .../featurizers_ops/stringtransformer_test.cc |   8 +-
 22 files changed, 1379 insertions(+), 367 deletions(-)
 create mode 100644 onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc
 create mode 100644 onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc
 create mode 100644 onnxruntime/featurizers_ops/cpu/min_max_scalar_transformer.cc
 create mode 100644 onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc
 create mode 100644 onnxruntime/featurizers_ops/cpu/robust_scalar_transformer.cc
 create mode 100644 onnxruntime/test/featurizers_ops/imputationmarkertransformer_test.cc
 create mode 100644 onnxruntime/test/featurizers_ops/labelencodertransfomer_test.cc
 rename onnxruntime/test/featurizers_ops/{maxabsscalerfeaturizer_test.cc => maxabsscalertransformer_test.cc} (79%)
 create mode 100644 onnxruntime/test/featurizers_ops/minmaxscalartransformer_test.cc
 create mode 100644 onnxruntime/test/featurizers_ops/missingdummiestransfomer_test.cc
 create mode 100644 onnxruntime/test/featurizers_ops/robustscalartransformer_test.cc

diff --git a/cgmanifest.json b/cgmanifest.json
index 782db27f7c8d0..e2b2c73a60a57 100644
--- a/cgmanifest.json
+++ b/cgmanifest.json
@@ -450,7 +450,7 @@
       {
          "component": {
             "git": {
-               "commitHash": "3f0f9802553944b75015aad098d856b2d17220df",
+               "commitHash": "a11f5002af58a03d5902b13ef65c84cedb499024",
                "repositoryUrl": "https://github.com/microsoft/FeaturizersLibrary.git"
             },
             "type": "git"
diff --git a/cmake/external/featurizers.cmake b/cmake/external/featurizers.cmake
index ac7f2432db921..acc9b25c564d1 100644
--- a/cmake/external/featurizers.cmake
+++ b/cmake/external/featurizers.cmake
@@ -3,15 +3,14 @@
 # This source code should not depend on the onnxruntime and may be built independently
 
 set(featurizers_URL "https://github.com/microsoft/FeaturizersLibrary.git")
-set(featurizers_TAG "3f0f9802553944b75015aad098d856b2d17220df")
+set(featurizers_TAG "a11f5002af58a03d5902b13ef65c84cedb499024")
 
 set(featurizers_pref FeaturizersLibrary)
 set(featurizers_ROOT ${PROJECT_SOURCE_DIR}/external/${featurizers_pref})
 set(featurizers_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR}/external/${featurizers_pref})
 
-# Only due to GIT_CONFIG
-# Uncoment UPDATE_COMMAND if you work locally
-# on the featurizers so cmake does not undo your changes.
+# Windows required git config setting
+# and an a switch whether we link to MSVCRT statically or dynamically
 if (WIN32)
     ExternalProject_Add(featurizers_lib
             PREFIX ${featurizers_pref}
@@ -24,7 +23,6 @@ if (WIN32)
             SOURCE_SUBDIR src/Featurizers
             BINARY_DIR ${featurizers_BINARY_DIR}
             CMAKE_ARGS -Dfeaturizers_MSVC_STATIC_RUNTIME=${onnxruntime_MSVC_STATIC_RUNTIME}
-#            UPDATE_COMMAND ""
             INSTALL_COMMAND ""
         )
 else()
@@ -37,7 +35,6 @@ else()
             SOURCE_SUBDIR src/Featurizers
             BINARY_DIR ${featurizers_BINARY_DIR}
             CMAKE_ARGS -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-#            UPDATE_COMMAND ""
             INSTALL_COMMAND ""
         )
 endif()
diff --git a/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc b/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
index d6db03b3f6c30..5c12ef18bce8f 100644
--- a/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
+++ b/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
@@ -34,7 +34,12 @@ using ONNX_NAMESPACE::OPTIONAL;
 // Forward declarations
 static void RegisterCatImputerFeaturizerVer1();
 static void RegisterDateTimeFeaturizerVer1();
+static void RegisterImputationMarkerFeaturizerVer1();
+static void RegisterLabelEncoderFeaturizerVer1();
 static void RegisterMaxAbsScalarFeaturizerVer1();
+static void RegisterMinMaxScalarFeaturizerVer1();
+static void RegisterMissingDummiesFeaturizerVer1();
+static void RegisterRobustScalarFeaturizerVer1();
 static void RegisterStringFeaturizerVer1();
 
 // ----------------------------------------------------------------------
@@ -43,7 +48,12 @@ static void RegisterStringFeaturizerVer1();
 void RegisterMSFeaturizersSchemas() {
   RegisterCatImputerFeaturizerVer1();
   RegisterDateTimeFeaturizerVer1();
+  RegisterImputationMarkerFeaturizerVer1();
+  RegisterLabelEncoderFeaturizerVer1();
   RegisterMaxAbsScalarFeaturizerVer1();
+  RegisterMinMaxScalarFeaturizerVer1();
+  RegisterMissingDummiesFeaturizerVer1();
+  RegisterRobustScalarFeaturizerVer1();
   RegisterStringFeaturizerVer1();
 }
 
@@ -76,7 +86,7 @@ void RegisterCatImputerFeaturizerVer1() {
           0,
           "State",
           "State generated during training that is used for prediction",
-          "tensor(uint8)")
+          "T0")
       .Input(
           1,
           "Input",
@@ -87,6 +97,10 @@ void RegisterCatImputerFeaturizerVer1() {
           "Output",
           "No information is available",
           "T")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
       .TypeConstraint(
           "T",
           {"tensor(float)", "tensor(double)", "tensor(string)"},
@@ -94,10 +108,9 @@ void RegisterCatImputerFeaturizerVer1() {
       .TypeAndShapeInferenceFunction(
           [](ONNX_NAMESPACE::InferenceContext& ctx) {
             propagateElemTypeFromInputToOutput(ctx, 1, 0);
-            if (!hasNInputShapes(ctx, 1)) {
-              return;
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
             }
-            propagateShapeFromInputToOutput(ctx, 1, 0);
           });
 }
 
@@ -144,7 +157,7 @@ void RegisterDateTimeFeaturizerVer1() {
           0,
           "State",
           "State generated during training that is used for prediction",
-          "tensor(uint8)")
+          "T0")
       .Input(
           1,
           "Input",
@@ -171,6 +184,10 @@ void RegisterDateTimeFeaturizerVer1() {
       .Output(18, "dayOfWeekLabel", "No information available", "OutputT3")
       .Output(19, "holidayName", "No information available", "OutputT3")
       .Output(20, "isPaidTimeOff", "No information available", "OutputT1")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
       .TypeConstraint(
           "OutputT0",
           {"tensor(int32)"},
@@ -189,30 +206,156 @@ void RegisterDateTimeFeaturizerVer1() {
           "No information is available")
       .TypeAndShapeInferenceFunction(
           [](ONNX_NAMESPACE::InferenceContext& ctx) {
-            ctx.getOutputType(0)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_INT32);
-            ctx.getOutputType(1)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(2)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(3)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(4)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(5)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(6)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(7)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(8)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(9)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(10)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT16);
-            ctx.getOutputType(11)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT16);
-            ctx.getOutputType(12)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(13)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(14)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-            ctx.getOutputType(15)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_INT32);
-            ctx.getOutputType(16)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_STRING);
-            ctx.getOutputType(17)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_STRING);
-            ctx.getOutputType(18)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_STRING);
-            ctx.getOutputType(19)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_STRING);
-            ctx.getOutputType(20)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_UINT8);
-
-            for (size_t i = 0; i < ctx.getNumOutputs(); ++i) {
-              *ctx.getOutputType(i)->mutable_tensor_type()->mutable_shape() = ctx.getInputType(1)->tensor_type().shape();
+            const bool has_shape = hasInputShape(ctx, 1);
+            for (int output = 0; output < 21; ++output) {
+              switch (output) {
+                case 0:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, output);
+                  break;
+                case 1: // fall through
+                case 2:
+                case 3:
+                case 4:
+                case 5:
+                case 6:
+                case 7:
+                case 8:
+                case 9:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
+                  break;
+                case 10: // fall through
+                case 11:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT16, output);
+                  break;
+                case 12: // fall through
+                case 13:
+                case 14:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
+                  break;
+                case 15:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, output);
+                  break;
+                case 16:
+                case 17:
+                case 18:
+                case 19:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, output);
+                  break;
+                case 20:
+                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
+                  break;
+                default:
+                  assert(false);
+                  break;
+              }
+              if (has_shape) {
+                propagateShapeFromInputToOutput(ctx, 1, output);
+              }
+            }
+          });
+}
+
+void RegisterImputationMarkerFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        Returns true if the input is null, false if it is not.
+
+        C++-style pseudo signature:
+          bool execute(std::float_t const &value);
+          bool execute(std::double_t const &value);
+          template <typename T> bool execute(std::optional<T> const &value);
+
+        Examples:
+          3.0 -> false
+          NaN -> true
+          "foo" -> false
+          std::optional<std::string>() -> true
+          std::optional<std::string>("bar") -> false
+    )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(ImputationMarkerTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(
+          0,
+          "Output",
+          "No information is available",
+          "tensor(bool)")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(float)", "tensor(double)", "tensor(string)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_BOOL, 0);
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
+          });
+}
+
+void RegisterLabelEncoderFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        Returns a unique id for the input based on all values encountered during training.
+
+        C++-style pseudo signature:
+          template <typename T> std::uint32_t execute(T const &value);
+
+        Examples:
+          Assuming the training data of ["A", "B", "C"]...
+
+          execute("A") -> 1
+          execute("B") -> 2
+          execute("C") -> 3
+          execute("This value was not seen during training") -> 0
+    )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(LabelEncoderTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(
+          0,
+          "Output",
+          "No information is available",
+          "tensor(uint32)")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(int8)", "tensor(int16)", "tensor(int32)", "tensor(int64)", "tensor(uint8)", "tensor(uint16)", "tensor(uint32)", "tensor(uint64)", "tensor(float)", "tensor(double)", "tensor(bool)", "tensor(string)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT32, 0);
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
             }
           });
 }
@@ -242,7 +385,7 @@ void RegisterMaxAbsScalarFeaturizerVer1() {
           0,
           "State",
           "State generated during training that is used for prediction",
-          "tensor(uint8)")
+          "T0")
       .Input(
           1,
           "Input",
@@ -253,6 +396,10 @@ void RegisterMaxAbsScalarFeaturizerVer1() {
           "Output",
           "No information is available",
           "OutputT")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
       .TypeConstraint(
           "InputT",
           {"tensor(int8)", "tensor(int16)", "tensor(uint8)", "tensor(uint16)", "tensor(float)", "tensor(int32)", "tensor(int64)", "tensor(uint32)", "tensor(uint64)", "tensor(double)"},
@@ -269,16 +416,193 @@ void RegisterMaxAbsScalarFeaturizerVer1() {
                 input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT8 ||
                 input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT16 ||
                 input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_FLOAT) {
-              ctx.getOutputType(0)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_FLOAT);
+              propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_FLOAT, 0);
             } else if (input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_INT32 ||
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_INT64 ||
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT32 ||
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT64 ||
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_DOUBLE) {
-              ctx.getOutputType(0)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_DOUBLE);
+              propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_DOUBLE, 0);
+            } else {
+              fail_type_inference("input 1 is expected to have a accepted type");
+            }
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
             }
+          });
+}
+
+void RegisterMinMaxScalarFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        Scales input based on the scale that results from the minimum and maximum values encountered
+        during training.
+
+        C++-style pseudo signature:
+            template <typeanem T> std::double_t(T const &value);
+
+        Examples:
+          Given the training data [1, 2, 3, 4, 5];
+            min: 1
+            max: 5
+            scale (<max> - <min>): 4
+
+          execute(2) = 2 / 4
+          execute(20) = 20 / 4
+    )DOC";
 
-            *ctx.getOutputType(0)->mutable_tensor_type()->mutable_shape() = ctx.getInputType(1)->tensor_type().shape();
+  MS_FEATURIZERS_OPERATOR_SCHEMA(MinMaxScalarTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(
+          0,
+          "Output",
+          "No information is available",
+          "tensor(double)")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(int8)", "tensor(int16)", "tensor(int32)", "tensor(int64)", "tensor(uint8)", "tensor(uint16)", "tensor(uint32)", "tensor(uint64)", "tensor(float)", "tensor(double)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_DOUBLE, 0);
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
+          });
+}
+
+void RegisterMissingDummiesFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        Returns 1 if the input is null, 0 if it is not.
+
+        C++-style pseudo signature:
+            std::int8_t execute(std::float_t const &value);
+            std::int8_t execute(std::double_t const &value);
+            template <typename T> std::int8_t execute(T const &value);
+
+        Examples:
+          1.0 -> 0
+          NaN -> 1
+          "foo" -> 0
+          std::optional<std::string>() -> 1
+          std::optional<std::string>("bar") -> 0
+    )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(MissingDummiesTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(
+          0,
+          "Output",
+          "No information is available",
+          "tensor(int8)")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(float)", "tensor(double)", "tensor(string)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT8, 0);
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
+          });
+}
+
+void RegisterRobustScalarFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        MinMaxScalarEstimator + centering?
+
+        C++-style pseudo signature:
+            TODO
+
+        Examples:
+          TODO
+    )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(RobustScalarTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(
+          0,
+          "Output",
+          "No information is available",
+          "OutputT")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(int8)", "tensor(int16)", "tensor(uint8)", "tensor(uint16)", "tensor(float)", "tensor(int32)", "tensor(int64)", "tensor(uint32)", "tensor(uint64)", "tensor(double)"},
+          "No information is available")
+      .TypeConstraint(
+          "OutputT",
+          {"tensor(float)", "tensor(double)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            auto input_elem_type = ctx.getInputType(1)->tensor_type().elem_type();
+            if (input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_INT8 ||
+                input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_INT16 ||
+                input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT8 ||
+                input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT16 ||
+                input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_FLOAT) {
+              propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_FLOAT, 0);
+            } else if (input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_INT32 ||
+                       input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_INT64 ||
+                       input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT32 ||
+                       input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT64 ||
+                       input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_DOUBLE) {
+              ctx.getOutputType(0)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_DOUBLE);
+              propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_DOUBLE, 0);
+            } else {
+              fail_type_inference("input 1 is expected to have a accepted type");
+            }
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
           });
 }
 
@@ -302,7 +626,7 @@ void RegisterStringFeaturizerVer1() {
           0,
           "State",
           "State generated during training that is used for prediction",
-          "tensor(uint8)")
+          "T0")
       .Input(
           1,
           "Input",
@@ -313,6 +637,10 @@ void RegisterStringFeaturizerVer1() {
           "Output",
           "No information is available",
           "tensor(string)")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
       .TypeConstraint(
           "InputT",
           {"tensor(int8)", "tensor(int16)", "tensor(int32)", "tensor(int64)", "tensor(uint8)", "tensor(uint16)", "tensor(uint32)", "tensor(uint64)", "tensor(float)", "tensor(double)", "tensor(bool)", "tensor(string)"},
@@ -320,8 +648,7 @@ void RegisterStringFeaturizerVer1() {
       .TypeAndShapeInferenceFunction(
           [](ONNX_NAMESPACE::InferenceContext& ctx) {
             propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, 0);
-
-            *ctx.getOutputType(0)->mutable_tensor_type()->mutable_shape() = ctx.getInputType(1)->tensor_type().shape();
+            propagateShapeFromInputToOutput(ctx, 1, 0);
           });
 }
 
diff --git a/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc b/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc
index 7aa53d734e298..b2adb9108f8b9 100644
--- a/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc
@@ -3,6 +3,7 @@
 
 #include "core/common/common.h"
 #include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
 #include "core/framework/op_kernel.h"
 
 #include "Featurizers/CatImputerFeaturizer.h"
@@ -26,13 +27,9 @@ inline nonstd::optional<std::string> PreprocessOptional(std::string value) {
   return value.empty() ? nonstd::optional<std::string>() : nonstd::optional<std::string>(std::move(value));
 }
 
-template <typename T>
-class CatImputerTransformer final : public OpKernel {
- public:
-  explicit CatImputerTransformer(const OpKernelInfo& info) : OpKernel(info) {
-  }
-
-  Status Compute(OpKernelContext* ctx) const override {
+template<typename T>
+struct CatImputerTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
     // Create the transformer
     Microsoft::Featurizer::Featurizers::CatImputerTransformer<T> transformer(
         [ctx](void) {
@@ -57,40 +54,32 @@ class CatImputerTransformer final : public OpKernel {
     for (int64_t i = 0; i < length; ++i) {
       output_data[i] = transformer.execute(PreprocessOptional(input_data[i]));
     }
-
-    return Status::OK();
   }
 };
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    CatImputerTransformer,
-    kMSFeaturizersDomain,
-    1,
-    float,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("T", DataTypeImpl::GetTensorType<float>()),
-    CatImputerTransformer<float>);
+class CatImputerTransformer final : public OpKernel {
+ public:
+  explicit CatImputerTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    CatImputerTransformer,
-    kMSFeaturizersDomain,
-    1,
-    double,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("T", DataTypeImpl::GetTensorType<double>()),
-    CatImputerTransformer<double>);
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<CatImputerTransformerImpl, float, double, std::string> t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
+ONNX_OPERATOR_KERNEL_EX(
     CatImputerTransformer,
     kMSFeaturizersDomain,
     1,
-    string,
     kCpuExecutionProvider,
     KernelDefBuilder()
-        .TypeConstraint("T", DataTypeImpl::GetTensorType<std::string>()),
-    CatImputerTransformer<std::string>);
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(),
+                              DataTypeImpl::GetTensorType<double>(),
+                              DataTypeImpl::GetTensorType<std::string>()}),
+    CatImputerTransformer);
 
 }  // namespace featurizers
 }  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc b/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc
index 1ac583b0e5b76..0e25e58fac68e 100644
--- a/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc
@@ -115,7 +115,7 @@ ONNX_OPERATOR_KERNEL_EX(
     1,
     kCpuExecutionProvider,
     KernelDefBuilder()
-        .TypeConstraint("T", DataTypeImpl::GetTensorType<int8_t>())
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
         .TypeConstraint("T1", DataTypeImpl::GetTensorType<uint64_t>()),
     DateTimeTransformer);
 
diff --git a/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc b/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc
new file mode 100644
index 0000000000000..a0f36a943ff01
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc
@@ -0,0 +1,76 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/ImputationMarkerFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+inline float const& PreprocessOptional(float const& value) { return value; }
+inline double const& PreprocessOptional(double const& value) { return value; }
+inline nonstd::optional<std::string> PreprocessOptional(std::string value) {
+  return value.empty() ? nonstd::optional<std::string>() : nonstd::optional<std::string>(std::move(value));
+}
+
+template <typename InputT>
+struct ImputationMarkerTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::ImputationMarkerTransformer<InputT> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::ImputationMarkerTransformer<InputT>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* output_tensor(ctx->Output(0, input_tensor->Shape()));
+    bool* output_data(output_tensor->MutableData<bool>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      output_data[i] = transformer.execute(PreprocessOptional(input_data[i]));
+    }
+  }
+};
+
+class ImputationMarkerTransformer final : public OpKernel {
+ public:
+  explicit ImputationMarkerTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<ImputationMarkerTransformerImpl, float, double, std::string>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    ImputationMarkerTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>(),
+                                   DataTypeImpl::GetTensorType<std::string>()}),
+    ImputationMarkerTransformer);
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc b/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc
new file mode 100644
index 0000000000000..96ec25285133c
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc
@@ -0,0 +1,80 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/LabelEncoderFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+template <typename InputT>
+struct LabelEncoderTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::LabelEncoderTransformer<InputT> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::LabelEncoderTransformer<InputT>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* output_tensor(ctx->Output(0, input_tensor->Shape()));
+    std::uint32_t* output_data(output_tensor->MutableData<std::uint32_t>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      output_data[i] = transformer.execute(input_data[i]);
+    }
+  }
+};
+
+class LabelEncoderTransformer final : public OpKernel {
+ public:
+  explicit LabelEncoderTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<LabelEncoderTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
+                                int64_t, uint64_t, float, double, bool, std::string>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    LabelEncoderTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>(),
+                                   DataTypeImpl::GetTensorType<bool>(),
+                                   DataTypeImpl::GetTensorType<std::string>()}),
+    LabelEncoderTransformer);
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc b/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc
index 98b7845997e34..8fea4f9fb806e 100644
--- a/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc
@@ -3,6 +3,7 @@
 
 #include "core/common/common.h"
 #include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
 #include "core/framework/op_kernel.h"
 
 #include "Featurizers/MaxAbsScalarFeaturizer.h"
@@ -35,12 +36,8 @@ template <>
 struct OutputTypeMapper<double> { using type = double; };
 
 template <typename InputT>
-class MaxAbsScalarTransformer final : public OpKernel {
- public:
-  explicit MaxAbsScalarTransformer(const OpKernelInfo& info) : OpKernel(info) {
-  }
-
-  Status Compute(OpKernelContext* ctx) const override {
+struct MaxAbsScalarTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
     // Create the transformer
     Microsoft::Featurizer::Featurizers::MaxAbsScalarTransformer<InputT, typename OutputTypeMapper<InputT>::type> transformer(
         [ctx](void) {
@@ -65,110 +62,39 @@ class MaxAbsScalarTransformer final : public OpKernel {
     for (int64_t i = 0; i < length; ++i) {
       output_data[i] = transformer.execute(input_data[i]);
     }
-
-    return Status::OK();
   }
 };
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int8,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int8_t>()),
-    MaxAbsScalarTransformer<int8_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int16,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int16_t>()),
-    MaxAbsScalarTransformer<int16_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint8,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint8_t>()),
-    MaxAbsScalarTransformer<uint8_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint16,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint16_t>()),
-    MaxAbsScalarTransformer<uint16_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    float,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<float>()),
-    MaxAbsScalarTransformer<float>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int32,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int32_t>()),
-    MaxAbsScalarTransformer<int32_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int64,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int64_t>()),
-    MaxAbsScalarTransformer<int64_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint32,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint32_t>()),
-    MaxAbsScalarTransformer<uint32_t>);
+class MaxAbsScalarTransformer final : public OpKernel {
+ public:
+  explicit MaxAbsScalarTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    MaxAbsScalarTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint64,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint64_t>()),
-    MaxAbsScalarTransformer<uint64_t>);
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<MaxAbsScalarTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
+      int64_t, uint64_t, float, double> t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
+ONNX_OPERATOR_KERNEL_EX(
     MaxAbsScalarTransformer,
     kMSFeaturizersDomain,
     1,
-    double,
     kCpuExecutionProvider,
     KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<double>()),
-    MaxAbsScalarTransformer<double>);
-
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>()}),
+    MaxAbsScalarTransformer);
 }  // namespace featurizers
 }  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/min_max_scalar_transformer.cc b/onnxruntime/featurizers_ops/cpu/min_max_scalar_transformer.cc
new file mode 100644
index 0000000000000..281e9cd8099d2
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/min_max_scalar_transformer.cc
@@ -0,0 +1,78 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/MinMaxScalarFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+template <typename InputT>
+struct MinMaxScalarTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::MinMaxScalarTransformer<InputT> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::MinMaxScalarTransformer<InputT>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* output_tensor(ctx->Output(0, input_tensor->Shape()));
+    double* output_data(output_tensor->MutableData<double>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      output_data[i] = transformer.execute(input_data[i]);
+    }
+  }
+};
+
+class MinMaxScalarTransformer final : public OpKernel {
+ public:
+  explicit MinMaxScalarTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<MinMaxScalarTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
+                                int64_t, uint64_t, float, double>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    MinMaxScalarTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>()}),
+    MinMaxScalarTransformer);
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc b/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc
new file mode 100644
index 0000000000000..255bfd78d3cc3
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc
@@ -0,0 +1,77 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/MissingDummiesFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+inline float const& PreprocessOptional(float const& value) { return value; }
+inline double const& PreprocessOptional(double const& value) { return value; }
+inline nonstd::optional<std::string> PreprocessOptional(std::string value) {
+  return value.empty() ? nonstd::optional<std::string>() : nonstd::optional<std::string>(std::move(value));
+}
+
+template <typename InputT>
+struct MissingDummiesTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::MissingDummiesTransformer<InputT> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::MissingDummiesTransformer<InputT>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* output_tensor(ctx->Output(0, input_tensor->Shape()));
+    int8_t* output_data(output_tensor->MutableData<int8_t>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      output_data[i] = transformer.execute(PreprocessOptional(input_data[i]));
+    }
+  }
+};
+
+class MissingDummiesTransformer final : public OpKernel {
+ public:
+  explicit MissingDummiesTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<MissingDummiesTransformerImpl, float, double, std::string>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    MissingDummiesTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>(),
+                                   DataTypeImpl::GetTensorType<std::string>()}),
+    MissingDummiesTransformer);
+
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/robust_scalar_transformer.cc b/onnxruntime/featurizers_ops/cpu/robust_scalar_transformer.cc
new file mode 100644
index 0000000000000..728c9c710508f
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/robust_scalar_transformer.cc
@@ -0,0 +1,102 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/RobustScalarFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+template <typename T>
+struct OutputTypeMapper {};
+template <>
+struct OutputTypeMapper<int8_t> { using type = float; };
+template <>
+struct OutputTypeMapper<int16_t> { using type = float; };
+template <>
+struct OutputTypeMapper<uint8_t> { using type = float; };
+template <>
+struct OutputTypeMapper<uint16_t> { using type = float; };
+template <>
+struct OutputTypeMapper<float> { using type = float; };
+template <>
+struct OutputTypeMapper<int32_t> { using type = double; };
+template <>
+struct OutputTypeMapper<int64_t> { using type = double; };
+template <>
+struct OutputTypeMapper<uint32_t> { using type = double; };
+template <>
+struct OutputTypeMapper<uint64_t> { using type = double; };
+template <>
+struct OutputTypeMapper<double> { using type = double; };
+
+template <typename InputT>
+struct RobustScalarTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::RobustScalarTransformer<InputT, typename OutputTypeMapper<InputT>::type> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::RobustScalarTransformer<InputT, typename OutputTypeMapper<InputT>::type>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* output_tensor(ctx->Output(0, input_tensor->Shape()));
+    typename OutputTypeMapper<InputT>::type* output_data(output_tensor->MutableData<typename OutputTypeMapper<InputT>::type>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      output_data[i] = transformer.execute(input_data[i]);
+    }
+  }
+};
+
+class RobustScalarTransformer final : public OpKernel {
+ public:
+  explicit RobustScalarTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<RobustScalarTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
+                                int64_t, uint64_t, float, double>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    RobustScalarTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>()}),
+    RobustScalarTransformer);
+
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/string_transformer.cc b/onnxruntime/featurizers_ops/cpu/string_transformer.cc
index 8f719552dd309..6ee65b8e2633d 100644
--- a/onnxruntime/featurizers_ops/cpu/string_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/string_transformer.cc
@@ -3,6 +3,7 @@
 
 #include "core/common/common.h"
 #include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
 #include "core/framework/op_kernel.h"
 
 #include "Featurizers/StringFeaturizer.h"
@@ -12,12 +13,8 @@ namespace onnxruntime {
 namespace featurizers {
 
 template <typename InputT>
-class StringTransformer final : public OpKernel {
- public:
-  explicit StringTransformer(const OpKernelInfo& info) : OpKernel(info) {
-  }
-
-  Status Compute(OpKernelContext* ctx) const override {
+struct StringTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
     // Create the transformer
     Microsoft::Featurizer::Featurizers::StringTransformer<InputT> transformer(
         [ctx](void) {
@@ -42,130 +39,43 @@ class StringTransformer final : public OpKernel {
     for (int64_t i = 0; i < length; ++i) {
       output_data[i] = transformer.execute(input_data[i]);
     }
-
-    return Status::OK();
   }
 };
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int8,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int8_t>()),
-    StringTransformer<int8_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int16,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int16_t>()),
-    StringTransformer<int16_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int32,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int32_t>()),
-    StringTransformer<int32_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    int64,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<int64_t>()),
-    StringTransformer<int64_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint8,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint8_t>()),
-    StringTransformer<uint8_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint16,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint16_t>()),
-    StringTransformer<uint16_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint32,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint32_t>()),
-    StringTransformer<uint32_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    uint64,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<uint64_t>()),
-    StringTransformer<uint64_t>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    float,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<float>()),
-    StringTransformer<float>);
-
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    double,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<double>()),
-    StringTransformer<double>);
+class StringTransformer final : public OpKernel {
+ public:
+  explicit StringTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
-    StringTransformer,
-    kMSFeaturizersDomain,
-    1,
-    bool,
-    kCpuExecutionProvider,
-    KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<bool>()),
-    StringTransformer<bool>);
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<StringTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
+                                int64_t, uint64_t, float, double, bool, std::string>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
 
-ONNX_OPERATOR_TYPED_KERNEL_EX(
+ONNX_OPERATOR_KERNEL_EX(
     StringTransformer,
     kMSFeaturizersDomain,
     1,
-    string,
     kCpuExecutionProvider,
     KernelDefBuilder()
-        .TypeConstraint("InputT", DataTypeImpl::GetTensorType<std::string>()),
-    StringTransformer<std::string>);
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>(),
+                                   DataTypeImpl::GetTensorType<bool>(),
+                                   DataTypeImpl::GetTensorType<std::string>()}),
+    StringTransformer);
 
 }  // namespace featurizers
 }  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc b/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
index 4acd5432ec22f..5f246c894b2b9 100644
--- a/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
+++ b/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
@@ -10,61 +10,28 @@ namespace onnxruntime {
 namespace featurizers {
 
 // Forward declarations
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, float, CatImputerTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, double, CatImputerTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, string, CatImputerTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, CatImputerTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, DateTimeTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int8, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int16, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint8, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint16, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, float, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int32, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int64, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint32, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint64, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, double, MaxAbsScalarTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int8, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int16, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int32, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int64, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint8, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint16, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint32, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint64, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, float, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, double, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, bool, StringTransformer);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, string, StringTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, ImputationMarkerTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, LabelEncoderTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MaxAbsScalarTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MinMaxScalarTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MissingDummiesTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, RobustScalarTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, StringTransformer);
 
 Status RegisterCpuMSFeaturizersKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, float, CatImputerTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, double, CatImputerTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, string, CatImputerTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, DateTimeTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int8, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int16, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint8, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint16, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, float, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int32, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int64, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint32, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint64, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, double, MaxAbsScalarTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int8, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int16, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int32, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, int64, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint8, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint16, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint32, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, uint64, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, float, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, double, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, bool, StringTransformer)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, string, StringTransformer)>};
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, CatImputerTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, DateTimeTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, ImputationMarkerTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, LabelEncoderTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MaxAbsScalarTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MinMaxScalarTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MissingDummiesTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, RobustScalarTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, StringTransformer)>,
+  };
 
   for (auto& function_table_entry : function_table) {
     ORT_RETURN_IF_ERROR(kernel_registry.Register(function_table_entry()));
diff --git a/onnxruntime/test/featurizers_ops/categoryimputer_test.cc b/onnxruntime/test/featurizers_ops/categoryimputer_test.cc
index d803b06879cce..5dc50cc245de2 100644
--- a/onnxruntime/test/featurizers_ops/categoryimputer_test.cc
+++ b/onnxruntime/test/featurizers_ops/categoryimputer_test.cc
@@ -11,7 +11,7 @@ namespace dft = Microsoft::Featurizer::Featurizers;
 namespace onnxruntime {
 namespace test {
 
-TEST(CategoryImputer, Float_values) {
+TEST(FeaturizersTests, CategoryImputer_float_values) {
 
   OpTester test("CatImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
   
@@ -20,15 +20,15 @@ TEST(CategoryImputer, Float_values) {
   test.AddInput<uint8_t>("State", {8}, {1, 0, 0, 0, 0, 0, 192, 63});
 
   // We are adding a scalar Tensor in this instance
-  test.AddInput<float_t>("Input", {5}, {1, std::nanf("1"), std::nanf("1"), 2, std::nanf("1")});
+  test.AddInput<float>("Input", {5}, {1.f, std::nanf("1"), std::nanf("1"), 2.f, std::nanf("1")});
 
   // Expected output.
-  test.AddOutput<float_t>("Output", {5}, {1, 1.5, 1.5, 2, 1.5});
+  test.AddOutput<float>("Output", {5}, {1.f, 1.5f, 1.5f, 2.f, 1.5f});
 
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(CategoryImputer, Double_values) {
+TEST(FeaturizersTests, CategoryImputer_double_values) {
   OpTester test("CatImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // State from when the transformer was trained. Corresponds to a
@@ -36,15 +36,15 @@ TEST(CategoryImputer, Double_values) {
   test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 248, 63});
 
   // We are adding a scalar Tensor in this instance
-  test.AddInput<double_t>("Input", {5}, {1, std::nan("1"), std::nan("1"), 2, std::nan("1")});
+  test.AddInput<double>("Input", {5}, {1., std::nan("1"), std::nan("1"), 2., std::nan("1")});
 
   // Expected output.
-  test.AddOutput<double_t>("Output", {5}, {1, 1.5, 1.5, 2, 1.5});
+  test.AddOutput<double>("Output", {5}, {1., 1.5, 1.5, 2., 1.5});
 
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(CategoryImputer, String_values) {
+TEST(FeaturizersTests, CategoryImputer_string_values) {
   OpTester test("CatImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // State from when the transformer was trained. Corresponds to a
diff --git a/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc b/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc
index 863fa7f1a2872..1044667fd54be 100644
--- a/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc
+++ b/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc
@@ -13,7 +13,7 @@ using SysClock = std::chrono::system_clock;
 namespace onnxruntime {
 namespace test {
 
-TEST(DateTimeTransformer, Past_1976_Nov_17__12_27_04) {
+TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17_12_27_04) {
   const time_t date = 217081624;
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
@@ -74,7 +74,7 @@ TEST(DateTimeTransformer, Past_1976_Nov_17__12_27_04) {
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(DateTimeTransformer, Past_1976_Nov_17__12_27_05) {
+TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17_12_27_05) {
   const time_t date = 217081625;
 
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
@@ -135,7 +135,7 @@ TEST(DateTimeTransformer, Past_1976_Nov_17__12_27_05) {
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(DateTimeTransformer, Past_1976_Nov_17__12_27_05_and_Past_1976_Nov_17__12_27_04) {
+TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17__12_27_05_and_past_1976_nov_17_12_27_04) {
   const time_t date1 = 217081625;
   const time_t date2 = 217081624;
 
@@ -223,7 +223,7 @@ TEST(DateTimeTransformer, Past_1976_Nov_17__12_27_05_and_Past_1976_Nov_17__12_27
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(DateTimeTransformer, Future_2025_June_30) {
+TEST(FeaturizersTests, DateTimeTransformer_future_2025_june_30) {
   const time_t date = 1751241600;
 
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
diff --git a/onnxruntime/test/featurizers_ops/imputationmarkertransformer_test.cc b/onnxruntime/test/featurizers_ops/imputationmarkertransformer_test.cc
new file mode 100644
index 0000000000000..df83269ead6d3
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/imputationmarkertransformer_test.cc
@@ -0,0 +1,146 @@
+// ----------------------------------------------------------------------
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License
+// ----------------------------------------------------------------------
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Featurizers/ImputationMarkerFeaturizer.h"
+
+namespace ft = Microsoft::Featurizer;
+
+namespace onnxruntime {
+namespace test {
+
+template<typename T>
+std::vector<uint8_t> GetStream () {
+  ft::Archive ar;
+  ft::Featurizers::ImputationMarkerTransformer<T> inst;
+  inst.save(ar);
+  return ar.commit();
+}
+
+//TEST (FeaturizersTests, ImputationMarker_int8) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<int8_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<int8_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_uint8) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<uint8_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<uint8_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_int16) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<int16_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<int16_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_uint16) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<uint16_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<uint16_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_int32) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<int32_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<int32_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_uint32) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<uint32_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<uint32_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_int64) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<int64_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<int64_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+//
+//TEST(FeaturizersTests, ImputationMarker_uint64) {
+//  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+//  auto stream = GetStream<uint64_t>();
+//  auto dim = static_cast<int64_t>(stream.size());
+//
+//  test.AddInput<uint8_t>("State", {dim}, stream);
+//  test.AddInput<uint64_t>("Input", {1}, {25});
+//  test.AddOutput<bool>("Output", {1}, {false});
+//  test.Run(OpTester::ExpectResult::kExpectSuccess);
+//}
+
+TEST(FeaturizersTests, ImputationMarker_float) {
+  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  auto stream = GetStream<float>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<float>("Input", {2}, {2.5f, std::numeric_limits<float>::quiet_NaN()});
+  test.AddOutput<bool>("Output", {2}, {false, true});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, ImputationMarker_double) {
+  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  auto stream = GetStream<double>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<double>("Input", {2}, {2.5, std::numeric_limits<double>::quiet_NaN()});
+  test.AddOutput<bool>("Output", {2}, {false, true});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, ImputationMarker_string) {
+  OpTester test("ImputationMarkerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  auto stream = GetStream<std::string>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<std::string>("Input", {2}, {"hello", ""});
+  test.AddOutput<bool>("Output", {2}, {false, true});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+}
+}
diff --git a/onnxruntime/test/featurizers_ops/labelencodertransfomer_test.cc b/onnxruntime/test/featurizers_ops/labelencodertransfomer_test.cc
new file mode 100644
index 0000000000000..32f1c380a6744
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/labelencodertransfomer_test.cc
@@ -0,0 +1,110 @@
+// ----------------------------------------------------------------------
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License
+// ----------------------------------------------------------------------
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Archive.h"
+#include "Featurizers/LabelEncoderFeaturizer.h"
+#include "Featurizers/TestHelpers.h"
+
+namespace ft = Microsoft::Featurizer;
+
+namespace onnxruntime {
+namespace test {
+
+template <typename KeyT, typename IndexT>
+using IndexMap = std::unordered_map<KeyT, IndexT>;
+
+template <typename InputType>
+std::vector<uint8_t> GetStream(const IndexMap<InputType, uint32_t>& map, bool allowMissingValues) {
+  ft::Archive ar;
+  using TransType = ft::Featurizers::LabelEncoderTransformer<InputType>;
+  TransType inst(map, allowMissingValues);
+  inst.save(ar);
+  return ar.commit();
+}
+
+TEST(FeaturizersTests, LabelEncodeTransformer_uint32) {
+  OpTester test("LabelEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  using InputType = uint32_t;
+
+  IndexMap<InputType, uint32_t> index_map = {
+      {11, 2}, {8, 0}, {10, 1}, {15, 3}, {20, 5}};
+
+  auto stream = GetStream<InputType>(index_map, false);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {5}, {11, 8, 10, 15, 20});
+  test.AddOutput<uint32_t>("Output", {5}, {2, 0, 1, 3, 5});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, LabelEncodeTransformer_string) {
+  OpTester test("LabelEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  using InputType = std::string;
+
+  IndexMap<InputType, uint32_t> index_map = {
+      {"orange", 5}, {"apple", 0}, {"grape", 3}, {"carrot", 5}, {"peach", 5}, {"banana", 1}};
+
+  auto stream = GetStream<InputType>(index_map, false);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {3}, {"banana", "grape", "apple"});
+  test.AddOutput<uint32_t>("Output", {3}, {1, 3, 0});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, LabelEncodeTransformer_string_nothrow) {
+  OpTester test("LabelEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  using InputType = std::string;
+
+  // when an inference data is not seen before, in the non-throw mode, the featurizer should generate 0
+  // hello is not seen before among fruits
+  IndexMap<InputType, uint32_t> index_map = {
+      {"banana", 1},
+      {"apple", 2},
+      {"grape", 3},
+      {"carrot", 4},
+      {"peach", 5},
+      {"orange", 6}};
+
+  auto stream = GetStream<InputType>(index_map, true);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {3}, {"banana", "grape", "hello"});
+  // The transformer will add 1 to each of the output for the missing input
+  test.AddOutput<uint32_t>("Output", {3}, {2, 4, 0});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, LabelEncodeTransformer_string_throw) {
+  OpTester test("LabelEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  using InputType = std::string;
+
+  // when an inference data is not seen before, in the non-throw mode, the featurizer should generate 0
+  // hello is not seen before among fruits
+  IndexMap<InputType, uint32_t> index_map = {
+      {"banana", 1},
+      {"apple", 2},
+      {"grape", 3},
+      {"carrot", 4},
+      {"peach", 5},
+      {"orange", 6}};
+
+  auto stream = GetStream<InputType>(index_map, false);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {4}, {"banana", "grape", "apple", "hello"});
+  test.AddOutput<uint32_t>("Output", {4}, {1, 3, 2, 0});
+  test.Run(OpTester::ExpectResult::kExpectFailure, "'input' was not found");
+}
+
+}  // namespace test
+}  // namespace onnxruntime
diff --git a/onnxruntime/test/featurizers_ops/maxabsscalerfeaturizer_test.cc b/onnxruntime/test/featurizers_ops/maxabsscalertransformer_test.cc
similarity index 79%
rename from onnxruntime/test/featurizers_ops/maxabsscalerfeaturizer_test.cc
rename to onnxruntime/test/featurizers_ops/maxabsscalertransformer_test.cc
index a871eec3efe26..3bc607f5289fa 100644
--- a/onnxruntime/test/featurizers_ops/maxabsscalerfeaturizer_test.cc
+++ b/onnxruntime/test/featurizers_ops/maxabsscalertransformer_test.cc
@@ -11,7 +11,7 @@ namespace dft = Microsoft::Featurizer::Featurizers;
 namespace onnxruntime {
 namespace test {
 
-TEST(MaxAbsScaler, Int8_values) {
+TEST(FeaturizersTests, MaxAbsScaler_int8_values) {
 
   OpTester test("MaxAbsScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
   
@@ -23,12 +23,12 @@ TEST(MaxAbsScaler, Int8_values) {
   test.AddInput<int8_t>("X", {5}, {-4,3,0,2,-1});
 
   // Expected output.
-  test.AddOutput<float_t>("ScaledValues", {5}, {-1,.75,0,.5,-.25});
+  test.AddOutput<float>("ScaledValues", {5}, {-1.f,.75f,0.f,.5f,-.25f});
 
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(MaxAbsScaler, Double_values) {
+TEST(FeaturizersTests, MaxAbsScaler_double_values) {
   OpTester test("MaxAbsScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // State from when the transformer was trained. Corresponds to Version 1 and a
@@ -36,10 +36,10 @@ TEST(MaxAbsScaler, Double_values) {
   test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 16, 64});
 
   // We are adding a scalar Tensor in this instance
-  test.AddInput<double_t>("X", {5}, {-4, 3, 0, 2, -1});
+  test.AddInput<double>("X", {5}, {-4, 3, 0, 2, -1});
 
   // Expected output.
-  test.AddOutput<double_t>("ScaledValues", {5}, {-1, .75, 0, .5, -.25});
+  test.AddOutput<double>("ScaledValues", {5}, {-1, .75, 0, .5, -.25});
 
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
diff --git a/onnxruntime/test/featurizers_ops/minmaxscalartransformer_test.cc b/onnxruntime/test/featurizers_ops/minmaxscalartransformer_test.cc
new file mode 100644
index 0000000000000..4551616b5d83f
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/minmaxscalartransformer_test.cc
@@ -0,0 +1,60 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Featurizers/MinMaxScalarFeaturizer.h"
+
+namespace onnxruntime {
+namespace test {
+
+TEST(FeaturizersTests, MinMaxScalarTransformer_int8) {
+  OpTester test("MinMaxScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {6}, {1, 0, 0, 0, 1, 9});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {1}, {15});
+
+  // Expected output.
+  test.AddOutput<double>("?2", {1}, {1.75});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+TEST(FeaturizersTests, MinMaxScalarTransformer_float_t) {
+  OpTester test("MinMaxScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 128, 191, 0, 0, 128, 63});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<float>("?1", {1}, {2.f});
+
+  // Expected output.
+  test.AddOutput<double>("?2", {1}, {1.5});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, MinMaxScalarTransformer_only_one_input) {
+  OpTester test("MinMaxScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {6}, {1, 0, 0, 0, 255, 255});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {1}, {2});
+
+  // Expected output.
+  test.AddOutput<double>("?2", {1}, {0});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+}
+}
\ No newline at end of file
diff --git a/onnxruntime/test/featurizers_ops/missingdummiestransfomer_test.cc b/onnxruntime/test/featurizers_ops/missingdummiestransfomer_test.cc
new file mode 100644
index 0000000000000..a457b4a5f4f43
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/missingdummiestransfomer_test.cc
@@ -0,0 +1,58 @@
+// ----------------------------------------------------------------------
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License
+// ----------------------------------------------------------------------
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Archive.h"
+#include "Featurizers/MissingDummiesFeaturizer.h"
+
+namespace ft = Microsoft::Featurizer;
+
+namespace onnxruntime {
+namespace test {
+
+template <typename T>
+std::vector<uint8_t> GetStream() {
+  ft::Archive ar;
+  ft::Featurizers::MissingDummiesTransformer<T> inst;
+  inst.save(ar);
+  return ar.commit();
+}
+
+TEST(FeaturizersTests, MissingDummiesTransformer_float) {
+  OpTester test("MissingDummiesTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  auto stream = GetStream<float>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<float>("Input", {2}, {2.5f, std::numeric_limits<float>::quiet_NaN()});
+  test.AddOutput<int8_t>("Output", {2}, {0, 1});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, MissingDummiesTransformer_double) {
+  OpTester test("MissingDummiesTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  auto stream = GetStream<double>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<double>("Input", {2}, {2.5, std::numeric_limits<double>::quiet_NaN()});
+  test.AddOutput<int8_t>("Output", {2}, {0, 1});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+TEST(FeaturizersTests, MissingDummiesTransformer_string) {
+  OpTester test("MissingDummiesTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  auto stream = GetStream<std::string>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<std::string>("Input", {2}, {"hello", ""});
+  test.AddOutput<int8_t>("Output", {2}, {0, 1});
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+}
+}
diff --git a/onnxruntime/test/featurizers_ops/robustscalartransformer_test.cc b/onnxruntime/test/featurizers_ops/robustscalartransformer_test.cc
new file mode 100644
index 0000000000000..60f809324e95b
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/robustscalartransformer_test.cc
@@ -0,0 +1,109 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Featurizers/RobustScalarFeaturizer.h"
+
+namespace onnxruntime {
+namespace test {
+
+TEST(FeaturizersTests, RobustScalarTransformer_default_with_centering) {
+  OpTester test("RobustScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 160, 64, 0, 0, 128, 64});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {5}, {1, 3, 5, 7, 9});
+
+  // Expected output.
+  test.AddOutput<float>("?2", {5}, {-1.0f,-0.5f, 0.0f, 0.5f, 1.0f});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+TEST(FeaturizersTests, RobustScalarTransformer_default_no_centering) {
+  OpTester test("RobustScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 64});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {5}, {1, 3, 5, 7, 9});
+
+  // Expected output.
+  test.AddOutput<float>("?2", {5}, {0.25f, 0.75f, 1.25f, 1.75f, 2.25f});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+TEST(FeaturizersTests, RobustScalarTransformer_default_no_centering_zero_scale) {
+  OpTester test("RobustScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {3}, {10, 10, 10});
+
+  // Expected output.
+  test.AddOutput<float>("?2", {3}, {10.f, 10.f, 10.f});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+TEST(FeaturizersTests, RobustScalarTransformer_default_with_centering_no_scaling) {
+  OpTester test("RobustScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 160, 64, 0, 0, 128, 63});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {5}, {1, 3, 5, 7, 9});
+
+  // Expected output.
+  test.AddOutput<float>("?2", {5}, {-4.f, -2.f, 0.f, 2.f, 4.f});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+TEST(FeaturizersTests, RobustScalarTransformer_default_with_centering_custom_scaling) {
+  OpTester test("RobustScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 160, 64, 0, 0, 0, 65});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {5}, {1, 3, 5, 7, 9});
+
+  // Expected output.
+  test.AddOutput<float>("?2", {5}, {-0.5f, -0.25f, 0.f, 0.25f, 0.5f});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+TEST(FeaturizersTests, RobustScalarTransformer_default_no_centering_custom_scaling) {
+  OpTester test("RobustScalarTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  // Add state input
+  test.AddInput<uint8_t>("State", {12}, {1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 65});
+
+  // We are adding a scalar Tensor in this instance
+  test.AddInput<int8_t>("?1", {5}, {1, 3, 5, 7, 9});
+
+  // Expected output.
+  test.AddOutput<float>("?2", {5}, {0.125f, 0.375f, 0.625f, 0.875f, 1.125f});
+
+  test.Run(OpTester::ExpectResult::kExpectSuccess);
+}
+
+
+}
+}
diff --git a/onnxruntime/test/featurizers_ops/stringtransformer_test.cc b/onnxruntime/test/featurizers_ops/stringtransformer_test.cc
index c2564d15c73e4..40af22313b9f7 100644
--- a/onnxruntime/test/featurizers_ops/stringtransformer_test.cc
+++ b/onnxruntime/test/featurizers_ops/stringtransformer_test.cc
@@ -11,7 +11,7 @@ namespace dft = Microsoft::Featurizer::Featurizers;
 namespace onnxruntime {
 namespace test {
 
-TEST(StringTransformer, Integer_values) {
+TEST(FeaturizersTests, StringTransformer_integer_values) {
   OpTester test("StringTransformer", 1, onnxruntime::kMSFeaturizersDomain);
   
   // State represents version 1
@@ -26,7 +26,7 @@ TEST(StringTransformer, Integer_values) {
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(StringTransformer, Double_values) {
+TEST(FeaturizersTests, StringTransformer_double_values) {
   OpTester test("StringTransformer", 1, onnxruntime::kMSFeaturizersDomain);
   
   // State represents version 1
@@ -41,7 +41,7 @@ TEST(StringTransformer, Double_values) {
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(StringTransformer, Bool_values) {
+TEST(FeaturizersTests, StringTransformer_bool_values) {
   OpTester test("StringTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // State represents version 1
@@ -56,7 +56,7 @@ TEST(StringTransformer, Bool_values) {
   test.Run(OpTester::ExpectResult::kExpectSuccess);
 }
 
-TEST(StringTransformer, String_values) {
+TEST(FeaturizersTests, StringTransformer_string_values) {
   OpTester test("StringTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // State represents version 1

From 72bdfc8cd47175635cb3cc8f055503a2d9706abb Mon Sep 17 00:00:00 2001
From: Yufeng Li <liyufeng1987@gmail.com>
Date: Mon, 6 Jan 2020 14:28:12 -0800
Subject: [PATCH 28/75] Implement a more stable softmax (#2715)

* Implement a more stable SoftMax
 e^x is represented as infinity if x is large enough, like 100.f. Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
A math transform as below is leveraged to get a stable softmax:
e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))

And for convenience, force max to 0.f if all xi are negative
---
 onnxruntime/contrib_ops/cpu/bert/attention.cc | 14 ++-
 .../contrib_ops/cuda/bert/attention_impl.cu   | 89 ++++++++++++++-----
 2 files changed, 77 insertions(+), 26 deletions(-)

diff --git a/onnxruntime/contrib_ops/cpu/bert/attention.cc b/onnxruntime/contrib_ops/cpu/bert/attention.cc
index 2027dd1da0ddc..9ec7e849c0cb6 100644
--- a/onnxruntime/contrib_ops/cpu/bert/attention.cc
+++ b/onnxruntime/contrib_ops/cpu/bert/attention.cc
@@ -238,8 +238,18 @@ Status Attention<T>::Compute(OpKernelContext* context) const {
       float* x = reinterpret_cast<T*>(scratch_data) + j * D;
       float* y = x;
 
-      for (int i = 0; i < D; i++)
-        y[i] = expf(x[i]);
+      // e^x is represented as infinity if x is large enough, like 100.f.
+      // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
+      // a math transform as below is leveraged to get a stable softmax:
+      // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
+      // And for convenience, force max to 0.f if all xi are negative
+      float max = 0.f;
+      for (int i = 0; i < D; i++) {
+        if (max < x[i]) max = x[i];
+      }
+      for (int i = 0; i < D; i++) {
+        y[i] = expf(x[i] - max);
+      }
 
       double sum = 0.0;
 
diff --git a/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu b/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu
index 53c79598ab5b7..5c72a1d53e316 100644
--- a/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu
+++ b/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu
@@ -54,84 +54,125 @@ size_t GetAttentionWorkspaceSize(size_t element_size, int batch_size, int num_he
 }
 
 template <typename T, unsigned TPB>
-__device__ inline void Softmax(const int ld, const int last_valid, const T* input, T* output) {
+__device__ inline void Softmax(const int ld, const int num_valid, const T* input, T* output) {
   using BlockReduce = cub::BlockReduce<float, TPB>;
   __shared__ typename BlockReduce::TempStorage tmp_storage;
 
-  __shared__ float reverse_z;
+  __shared__ float sum_reverse_block;
+  __shared__ float max_block;
 
   float thread_data(0);
+
   const int offset = (blockIdx.y * gridDim.x + blockIdx.x) * ld;
-  for (int i = threadIdx.x; i < last_valid; i += TPB) {
+  for (int i = threadIdx.x; i < num_valid; i += TPB) {
+    const int index = offset + i;
+    if (thread_data < float(input[index])) {
+      thread_data = float(input[index]);
+    }
+  }
+
+  // e^x is represented as infinity if x is large enough, like 100.f.
+  // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
+  // a math transform as below is leveraged to get a stable softmax:
+  // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
+  // And for convenience, force max to 0.f if all xi are negative
+  const auto max = BlockReduce(tmp_storage).Reduce(thread_data, cub::Max());
+
+  // Store max value
+  if (threadIdx.x == 0) {
+    max_block = max;
+  }
+  __syncthreads();
+
+  for (int i = threadIdx.x; i < num_valid; i += TPB) {
     const int index = offset + i;
     const float val = input[index];
-    thread_data += expf(val);
+    thread_data += expf(val - max_block);
   }
 
-  cub::Sum sum;
-  const auto z = BlockReduce(tmp_storage).Reduce(thread_data, sum);
+  const auto sum = BlockReduce(tmp_storage).Reduce(thread_data, cub::Sum());
   if (threadIdx.x == 0) {
-    reverse_z = 1.f / z;
+    sum_reverse_block = 1.f / sum;
   }
   __syncthreads();
 
   for (int i = threadIdx.x; i < ld; i += TPB) {
     const int index = offset + i;
-    const float val = (i < last_valid) ? expf(float(input[index])) * reverse_z : 0.f;
+    const float val = (i < num_valid) ? expf(float(input[index]) - max_block) * sum_reverse_block : 0.f;
     output[index] = T(val);
   }
 }
 
 template <typename T, unsigned TPB>
-__device__ inline void SoftmaxSmall(const int ld, const int last_valid, const T* input, T* output) {
+__device__ inline void SoftmaxSmall(const int ld, const int num_valid, const T* input, T* output) {
   using BlockReduce = cub::BlockReduce<float, TPB>;
   __shared__ typename BlockReduce::TempStorage tmp_storage;
 
-  __shared__ float reverse_z;
+  __shared__ float sum_reverse_block;
+  __shared__ float max_block;
 
-  float thread_data(0);
   const int offset = (blockIdx.y * gridDim.x + blockIdx.x) * ld;
   const int index = offset + threadIdx.x;
-  if (threadIdx.x < last_valid) {
+
+  float thread_data(0);
+  if (threadIdx.x < num_valid) {
+    thread_data = input[index];
+  }
+
+  // e^x is represented as infinity if x is large enough, like 100.f.
+  // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
+  // a math transform as below is leveraged to get a stable softmax:
+  // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
+  // And for convenience, force max to 0.f if all xi are negative
+  const auto max = BlockReduce(tmp_storage).Reduce(thread_data, cub::Max(), num_valid);
+
+  // Store max value
+  if (threadIdx.x == 0) {
+    max_block = max;
+  }
+  __syncthreads();
+
+  if (threadIdx.x < num_valid) {
     const float val = input[index];
-    thread_data = expf(val);
+    thread_data = expf(val - max_block);
   }
 
-  cub::Sum sum;
-  const auto z = BlockReduce(tmp_storage).Reduce(thread_data, sum);
+  const auto sum = BlockReduce(tmp_storage).Reduce(thread_data, cub::Sum(), num_valid);
+
+  // Store max value
   if (threadIdx.x == 0) {
-    reverse_z = (1.f) / z;
+    sum_reverse_block = (1.f) / sum;
   }
   __syncthreads();
 
   if (threadIdx.x < ld) {
-    // this will be 0 for threadIdx.x >= last_valid
-    output[index] = T(thread_data * reverse_z);
+    // this will be 0 for threadIdx.x >= num_valid
+    output[index] = T(thread_data * sum_reverse_block);
   }
 }
 
 template <typename T, unsigned TPB>
 __global__ void MaskedSoftmaxKernelSmall(const int sequence_length, const int* mask_index, const T* input, T* output) {
-  __shared__ int last_valid;
+  __shared__ int num_valid;
 
   if (threadIdx.x == 0) {
-    last_valid = min(sequence_length, mask_index[blockIdx.y]);
+    num_valid = min(sequence_length, mask_index[blockIdx.y]);
   }
   __syncthreads();
 
-  SoftmaxSmall<T, TPB>(sequence_length, last_valid, input, output);
+  SoftmaxSmall<T, TPB>(sequence_length, num_valid, input, output);
 }
 
 template <typename T, unsigned TPB>
 __global__ void MaskedSoftmaxKernel(const int sequence_length, const int* mask_index, const T* input, T* output) {
-  __shared__ int last_valid;
+  __shared__ int num_valid;
 
   if (threadIdx.x == 0) {
-    last_valid = min(sequence_length, mask_index[blockIdx.y]);
+    num_valid = min(sequence_length, mask_index[blockIdx.y]);
   }
   __syncthreads();
 
-  Softmax<T, TPB>(sequence_length, last_valid, input, output);
+  Softmax<T, TPB>(sequence_length, num_valid, input, output);
 }
 
 template <typename T>

From f22bffe0f62460850efda82ee2cc38c93da6ca48 Mon Sep 17 00:00:00 2001
From: Maher Jendoubi <maher.jendoubi@gmail.com>
Date: Tue, 7 Jan 2020 17:32:13 +0100
Subject: [PATCH 29/75] Contributing: Fix a typo (#2784)

---
 csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
index 70d48877da9c5..c2a5920741d4d 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
@@ -51,7 +51,7 @@ public SessionOptions()
 
 #if USE_CUDA
         /// <summary>
-        /// A helper method to constuct a SessionOptions object for CUDA execution
+        /// A helper method to construct a SessionOptions object for CUDA execution
         /// </summary>
         /// <returns>A SessionsOptions() object configured for execution on deviceId=0</returns>
         public static SessionOptions MakeSessionOptionWithCudaProvider()
@@ -60,7 +60,7 @@ public static SessionOptions MakeSessionOptionWithCudaProvider()
         }
 
         /// <summary>
-        /// A helper method to constuct a SessionOptions object for CUDA execution
+        /// A helper method to construct a SessionOptions object for CUDA execution
         /// </summary>
         /// <param name="deviceId"></param>
         /// <returns>A SessionsOptions() object configured for execution on deviceId</returns>

From fdc0106f8309fc3061d54afa7caf1c38c0ef6d76 Mon Sep 17 00:00:00 2001
From: Andrews548 <32704142+Andrews548@users.noreply.github.com>
Date: Tue, 7 Jan 2020 18:35:18 +0200
Subject: [PATCH 30/75] ACL EP GEMM improvements (#2780)

When it is posible we use a fully connected layer instead of the gemm implementation.
This will let the library use the best implementation based on the input data.
---
 onnxruntime/core/providers/acl/math/gemm.h | 58 ++++++++--------------
 1 file changed, 20 insertions(+), 38 deletions(-)

diff --git a/onnxruntime/core/providers/acl/math/gemm.h b/onnxruntime/core/providers/acl/math/gemm.h
index d898ad17b1a32..c171837c6d108 100644
--- a/onnxruntime/core/providers/acl/math/gemm.h
+++ b/onnxruntime/core/providers/acl/math/gemm.h
@@ -21,6 +21,7 @@
 //NEON
 #include "arm_compute/runtime/NEON/functions/NEGEMM.h"
 #include "arm_compute/runtime/NEON/functions/NETranspose.h"
+#include "arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h"
 
 #undef GEMM_ACL
 #define CACHE_TRANSPOSED_DATA
@@ -29,7 +30,7 @@ namespace onnxruntime {
 namespace acl {
 
 typedef struct {
-  std::shared_ptr<arm_compute::NEGEMM> layer;
+  std::shared_ptr<arm_compute::IFunction> layer;
   std::shared_ptr<arm_compute::Tensor> a, b, c, d;
   std::shared_ptr<arm_compute::MemoryManagerOnDemand> mm_layer;
 } ACLNEGEMM;
@@ -51,7 +52,6 @@ class Gemm : public onnxruntime::Gemm<T> {
     ORT_ENFORCE(info.GetAttr<float>("beta", &beta_).IsOK());
   }
 
-#ifdef GEMM_ACL
   Status Compute(OpKernelContext* context) const override {
     const auto X = context->Input<Tensor>(0);
     const auto W = context->Input<Tensor>(1);
@@ -66,6 +66,8 @@ class Gemm : public onnxruntime::Gemm<T> {
     int64_t N = helper.N();
     auto Y = context->Output(0, TensorShape({M, N}));
 
+    bool FC = ((alpha_ == 1 && beta_ == 1) || (alpha_ == 1 && beta_ == 0));
+
     int64_t K = helper.K();
     LOGS_DEFAULT(VERBOSE) << "Gemm ACL:" << std::endl;
     if (X) LOGS_DEFAULT(VERBOSE) << "X " << X->Shape().ToString().c_str() << std::endl;
@@ -93,7 +95,7 @@ class Gemm : public onnxruntime::Gemm<T> {
       tGEMM.d->allocator()->init(arm_compute::TensorInfo(arm_compute::TensorShape(N, M), tGEMM.a->info()->format()));
 
       // transpose
-      if (trans_B_ == CblasTrans) {
+      if (!FC && trans_B_ == CblasTrans) {
         auto trans_layer = std::make_shared<arm_compute::NETranspose>();
         tGEMM.b->allocator()->init(arm_compute::TensorInfo(ACLTensorShape(W->Shape()), arm_compute::Format::F32));
 
@@ -113,13 +115,23 @@ class Gemm : public onnxruntime::Gemm<T> {
       }
 
       tGEMM.mm_layer = ACLCreateMemoryManager();
-      tGEMM.layer = std::make_shared<arm_compute::NEGEMM>(tGEMM.mm_layer);
 
-      // configure GEMM
-      tGEMM.layer->configure(tGEMM.a.get(), tGEMM.b.get(), tGEMM.c.get(), tGEMM.d.get(), alpha_, beta_, arm_compute::GEMMInfo());
+      if(FC) {
+        auto layer = std::make_shared<arm_compute::NEFullyConnectedLayer>(tGEMM.mm_layer);
+        layer->configure(tGEMM.a.get(), tGEMM.b.get(), (B != nullptr && beta_ != 0) ? tGEMM.c.get() : nullptr, tGEMM.d.get());
+        tGEMM.layer = std::move(layer);
+      } else {
+#ifdef GEMM_ACL
+        auto layer = std::make_shared<arm_compute::NEGEMM>(tGEMM.mm_layer);
+        layer->configure(tGEMM.a.get(), tGEMM.b.get(), (B != nullptr && beta_ != 0) ? tGEMM.c.get() : nullptr, tGEMM.d.get(), alpha_, beta_, arm_compute::GEMMInfo());
+        tGEMM.layer = std::move(layer);
+#else
+        return onnxruntime::Gemm<T>::Compute(context);
+#endif
+      }
 
       // non-transpose
-      if (trans_B_ != CblasTrans) {
+      if (FC || trans_B_ != CblasTrans) {
         const T* b_data = W->template Data<T>();
         ACLImportMemory(tGEMM.b->allocator(), (void*)b_data, W->Shape().Size() * 4);
       }
@@ -132,7 +144,7 @@ class Gemm : public onnxruntime::Gemm<T> {
       pGEMM = &it->second;
 
       // transpose
-      if (trans_B_ == CblasTrans) {
+      if (!FC && trans_B_ == CblasTrans) {
 #ifndef CACHE_TRANSPOSED_DATA
         auto trans_layer = std::make_shared<arm_compute::NETranspose>();
         pGEMM->b->allocator()->init(arm_compute::TensorInfo(ACLTensorShape(W->Shape()), arm_compute::Format::F32));
@@ -193,36 +205,6 @@ class Gemm : public onnxruntime::Gemm<T> {
     gemmLayers.erase(this);
   }
 
-#else
-
-  Status Compute(OpKernelContext* context) const override {
-    const auto X = context->Input<Tensor>(0);
-    const auto W = context->Input<Tensor>(1);
-    const auto B = context->Input<Tensor>(2);
-
-    GemmHelper helper(X->Shape(), trans_A_ != CblasNoTrans, W->Shape(), trans_B_ != CblasNoTrans, B->Shape());
-
-    if (!helper.State().IsOK())
-      return helper.State();
-
-    int64_t M = helper.M();
-    int64_t N = helper.N();
-    auto Y = context->Output(0, TensorShape({M, N}));
-
-    int64_t K = helper.K();
-    LOGS_DEFAULT(VERBOSE) << "Gemm CPU:" << std::endl;
-    if (X) LOGS_DEFAULT(VERBOSE) << "X " << X->Shape().ToString().c_str() << std::endl;
-    if (W) LOGS_DEFAULT(VERBOSE) << "W " << W->Shape().ToString().c_str() << std::endl;
-    if (B) LOGS_DEFAULT(VERBOSE) << "B " << B->Shape().ToString().c_str() << std::endl;
-    LOGS_DEFAULT(VERBOSE) << "Y " << Y->Shape().ToString().c_str() << std::endl;
-    LOGS_DEFAULT(VERBOSE) << "M " << (int)M << ", N " << (int)N << ", K " << (int)K << std::endl;
-    LOGS_DEFAULT(VERBOSE) << "Alfa " << alpha_ << ", Beta " << beta_ << std::endl;
-    LOGS_DEFAULT(VERBOSE) << std::endl;
-
-    return onnxruntime::Gemm<T>::Compute(context);
-  }
-#endif
-
  private:
   static thread_local std::map<OpKernel*, ACLNEGEMM> gemmLayers;
 

From 3e6f1836eb8e6e51602c07e2539b2df90505528f Mon Sep 17 00:00:00 2001
From: Andrews548 <32704142+Andrews548@users.noreply.github.com>
Date: Tue, 7 Jan 2020 18:42:03 +0200
Subject: [PATCH 31/75] ACL EP convolution improvements (#2774)

Added the optimized implementation for depthwise convolution for both ACL v19.02 and ACL 19.05.
Also the pointwise convolution seems to be more optimal in the CPU implementation so we opted for that instead.
---
 onnxruntime/core/providers/acl/nn/conv.cc | 95 ++++++++++++++++++-----
 onnxruntime/core/providers/acl/nn/conv.h  |  4 +-
 2 files changed, 79 insertions(+), 20 deletions(-)

diff --git a/onnxruntime/core/providers/acl/nn/conv.cc b/onnxruntime/core/providers/acl/nn/conv.cc
index c191e27ed9a80..a4a42c908d24f 100644
--- a/onnxruntime/core/providers/acl/nn/conv.cc
+++ b/onnxruntime/core/providers/acl/nn/conv.cc
@@ -24,8 +24,14 @@
 #include "arm_compute/runtime/NEON/functions/NEConvolutionLayer.h"
 #include "arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h"
 
+#ifdef ACL_1902
+#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.h"
+#else
+#include "arm_compute/runtime/NEON/functions/assembly/NEDepthwiseConvolutionAssemblyDispatch.h"
+#endif
+
 #define CONV_ACL
-#define DEPTHWISE_CPU
+#undef DEPTHWISE_CPU
 
 #define PREF_DIM 4
 
@@ -36,7 +42,7 @@ template <typename T>
 thread_local std::map<OpKernel*, ACLNEConv> Conv<T>::convLayers;
 
 template <typename T>
-arm_compute::TensorShape Conv<T>::ACLReshapeWeightsDepthwise(arm_compute::Tensor* kernel) {
+arm_compute::TensorShape Conv<T>::ACLReshapeWeightsDepthwise(arm_compute::Tensor* kernel) const {
   arm_compute::TensorShape shape = arm_compute::TensorShape(kernel->info()->tensor_shape());
   shape[2] = shape[2] * shape[3];
   shape[3] = 1;
@@ -49,6 +55,16 @@ template <typename T>
 Status Conv<T>::Compute(OpKernelContext* context) const {
   size_t num_inputs = OpKernel::Node().InputDefs().size();
 
+  ACLNEConv* pConv;
+  ConvLayersIterator it = Conv::convLayers.find((OpKernel*)this);
+  if (it != Conv::convLayers.end()) {
+    pConv = &it->second;
+    if(pConv->isDepthwiseCPU == true) {
+      Status s = onnxruntime::Conv<T>::Compute(context);
+      return s;
+    }
+  }
+
   const Tensor* X = context->Input<Tensor>(0);
   const Tensor* W = context->Input<Tensor>(1);
   const Tensor* B = num_inputs == 3 ? context->Input<Tensor>(2) : nullptr;
@@ -109,9 +125,8 @@ Status Conv<T>::Compute(OpKernelContext* context) const {
     ORT_NOT_IMPLEMENTED("Not implemented fused activation: ", conv_attrs_.activation);
   }
 
-  ACLNEConv* pConv;
-  ConvLayersIterator it = Conv::convLayers.find((OpKernel*)this);
   if (it == Conv::convLayers.end()) {
+
     auto mm_layer = ACLCreateMemoryManager();
 
     ACLNEConv tconv;
@@ -133,6 +148,7 @@ Status Conv<T>::Compute(OpKernelContext* context) const {
     const arm_compute::DataLayout data_layout = tconv.in->info()->data_layout();
     const int idx_channel = arm_compute::get_data_layout_dimension_index(data_layout, arm_compute::DataLayoutDimension::CHANNEL);
     bool isDepthwise = (1 == tconv.k->info()->tensor_shape()[idx_channel]);
+    tconv.isDepthwiseCPU = isDepthwise;
 
     std::vector<int64_t> aclStrides(2);
     aclStrides[0] = (strides.size() == 2) ? strides[1] : 1;
@@ -160,29 +176,71 @@ Status Conv<T>::Compute(OpKernelContext* context) const {
 
     arm_compute::PadStrideInfo aclPadStride = arm_compute::PadStrideInfo(aclStrides[0], aclStrides[1],
                                                                          aclPads[0], aclPads[1], aclPads[2], aclPads[3], arm_compute::DimensionRoundingType::FLOOR);
+    unsigned int aclDilation0 = (dilations.size() == 2) ? dilations[1] : 1;
 
     if (isDepthwise) {
 #ifdef DEPTHWISE_CPU
       Status s = onnxruntime::Conv<T>::Compute(context);
+      std::pair<ConvLayersIterator, bool> ret;
+      ret = Conv::convLayers.insert(std::pair<OpKernel*, ACLNEConv>((OpKernel*)this, tconv));
       return s;
 #else
-      auto layer = std::make_shared<arm_compute::NEDepthwiseConvolutionLayer>();
       tconv.k->info()->set_tensor_shape(ACLReshapeWeightsDepthwise(tconv.k.get()));
-      layer->configure(tconv.in.get(), tconv.k.get(), (B != nullptr) ? tconv.b.get() : nullptr, tconv.out.get(),
-                       aclPadStride, 1 /* depth multiplier */,
-                       acl_activ_enabled ? arm_compute::ActivationLayerInfo(acl_activ_func, conv_attrs_.alpha) : arm_compute::ActivationLayerInfo());
-      tconv.layer = std::move(layer);
+
+      // in the configure function for NEDepthwiseConvolutionLayer3x3, there is a separation based on the optimization
+#ifdef ACL_1902
+      bool optimizable =
+            arm_compute::NEDepthwiseConvolutionLayer3x3Kernel::is_optimized_execution_possible(tconv.in->info()->tensor_shape(),
+                                                                                               aclPadStride,
+                                                                                               tconv.in->info()->data_type(),
+                                                                                               1 /* depth multiplier */,
+                                                                                               tconv.in->info()->data_layout());
+#else
+      bool optimizable =
+            arm_compute::NEDepthwiseConvolutionAssemblyDispatch::is_optimized_supported(tconv.in->info(),
+                                                                                        tconv.k->info(),
+                                                                                        aclPadStride,
+                                                                                        1 /* depth multiplier */,
+                                                                                        arm_compute::Size2D(aclDilation0, dilations[0]));
+#endif
+      if(optimizable) {
+        //optimized depthwise convolution
+        auto layer = std::make_shared<arm_compute::NEDepthwiseConvolutionLayer3x3>();
+#ifdef ACL_1902
+        layer->configure(tconv.in.get(), tconv.k.get(), (B != nullptr) ? tconv.b.get() : nullptr, tconv.out.get(),
+                         aclPadStride, 1 /* depth multiplier */,
+                         acl_activ_enabled ? arm_compute::ActivationLayerInfo(acl_activ_func, conv_attrs_.alpha) : arm_compute::ActivationLayerInfo());
+#else
+        layer->configure(tconv.in.get(), tconv.k.get(), (B != nullptr) ? tconv.b.get() : nullptr, tconv.out.get(),
+                         aclPadStride, 1 /* depth multiplier */,
+                         acl_activ_enabled ? arm_compute::ActivationLayerInfo(acl_activ_func, conv_attrs_.alpha) : arm_compute::ActivationLayerInfo(),
+                         arm_compute::Size2D(aclDilation0, dilations[0]));
+#endif
+        tconv.layer = std::move(layer);
+        tconv.isDepthwiseCPU = false;
+      } else {
+        // cpu depthwise convolution
+        Status s = onnxruntime::Conv<T>::Compute(context);
+        std::pair<ConvLayersIterator, bool> ret;
+        ret = Conv::convLayers.insert(std::pair<OpKernel*, ACLNEConv>((OpKernel*)this, tconv));
+        return s;
+      }
 #endif
     } else {
-      unsigned int aclDilation0 = (dilations.size() == 2) ? dilations[1] : 1;
-
-      auto layer = std::make_shared<arm_compute::NEConvolutionLayer>(mm_layer);
-      layer->configure(tconv.in.get(), tconv.k.get(), (B != nullptr) ? tconv.b.get() : nullptr, tconv.out.get(),
-                       aclPadStride,
-                       arm_compute::WeightsInfo(), arm_compute::Size2D(aclDilation0, dilations[0]),
-                       acl_activ_enabled ? arm_compute::ActivationLayerInfo(acl_activ_func, conv_attrs_.alpha) : arm_compute::ActivationLayerInfo(),
-                       false, conv_attrs_.group);
-      tconv.layer = std::move(layer);
+      if(tconv.k->info()->tensor_shape()[0] == 1 && tconv.k->info()->tensor_shape()[1] == 1) {
+        //pointwise convolution
+        Status s = onnxruntime::Conv<T>::Compute(context);
+        return s;
+      } else {
+        //convolution
+        auto layer = std::make_shared<arm_compute::NEConvolutionLayer>(mm_layer);
+        layer->configure(tconv.in.get(), tconv.k.get(), (B != nullptr) ? tconv.b.get() : nullptr, tconv.out.get(),
+                         aclPadStride,
+                         arm_compute::WeightsInfo(), arm_compute::Size2D(aclDilation0, dilations[0]),
+                         acl_activ_enabled ? arm_compute::ActivationLayerInfo(acl_activ_func, conv_attrs_.alpha) : arm_compute::ActivationLayerInfo(),
+                         false, conv_attrs_.group);
+        tconv.layer = std::move(layer);
+      }
     }
 
     tconv.out->info()->set_format(tconv.in->info()->format());
@@ -224,6 +282,7 @@ Status Conv<T>::Compute(OpKernelContext* context) const {
     pConv->b->allocator()->free();
   pConv->out->allocator()->free();
 
+
   return Status::OK();
 }
 #else
diff --git a/onnxruntime/core/providers/acl/nn/conv.h b/onnxruntime/core/providers/acl/nn/conv.h
index 218fe9173dddd..e4e413e454349 100644
--- a/onnxruntime/core/providers/acl/nn/conv.h
+++ b/onnxruntime/core/providers/acl/nn/conv.h
@@ -30,7 +30,7 @@ typedef struct
   std::shared_ptr<arm_compute::Tensor> k;
   std::shared_ptr<arm_compute::Tensor> b;
   std::shared_ptr<arm_compute::Tensor> out;
-  bool isDeptwise;
+  bool isDepthwiseCPU;
 } ACLNEConv;
 
 typedef std::map<OpKernel*, ACLNEConv>::iterator ConvLayersIterator;
@@ -54,7 +54,7 @@ class Conv final : public onnxruntime::Conv<T> {
   ConvAttributes conv_attrs_;
   ACLExecutionProvider* provider_;
 
-  arm_compute::TensorShape ACLReshapeWeightsDepthwise(arm_compute::Tensor* kernel);
+  arm_compute::TensorShape ACLReshapeWeightsDepthwise(arm_compute::Tensor* kernel) const;
 };
 }  // namespace mkl_dnn
 }  // namespace onnxruntime

From ebfcad1c90940a69fd370ae80110d4b1aa6fc949 Mon Sep 17 00:00:00 2001
From: Hariharan Seshadri <shariharan91@gmail.com>
Date: Wed, 8 Jan 2020 18:42:22 +0530
Subject: [PATCH 32/75] Add script for release Nuget validation (#2719)

* Initial commit

* Nits

* Disable a test temporarily

* Change working directory

* Test

* Add download python step

* Test update

* More changes

* Fix space issue

* Fix

* Verify nuget signing

* Fix

* Spaces

* PR feedback

* Nit

* Fix

* Fix

* Remove temporary changes
---
 .../nuget/templates/cpu-mklml.yml             |   6 +
 .../templates/cpu-nocontribops-arm64.yml      |   7 +
 .../azure-pipelines/nuget/templates/cpu.yml   |   6 +
 .../azure-pipelines/nuget/templates/gpu.yml   |   6 +
 .../templates/validate-nuget.yml              |  17 +++
 tools/nuget/validate_nuget.py                 | 140 ++++++++++++++++++
 6 files changed, 182 insertions(+)
 create mode 100644 tools/ci_build/github/azure-pipelines/templates/validate-nuget.yml
 create mode 100644 tools/nuget/validate_nuget.py

diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml
index afee0189d90d4..9d1ff3e8673a7 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-mklml.yml
@@ -166,6 +166,12 @@ jobs:
       FolderPath: '$(Build.ArtifactStagingDirectory)'
       DoEsrp: ${{ parameters.DoEsrp }}
 
+  - template: ../../templates/validate-nuget.yml
+    parameters:
+      NugetPath: '$(Build.ArtifactStagingDirectory)'
+      PlatformsSupported: 'win-x64,linux-x64,osx-x64'
+      VerifyNugetSigning: ${{ parameters.DoEsrp }}
+
   - task: PublishPipelineArtifact@0
     displayName: 'Publish Pipeline NuGet Artifact'
     inputs:
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml
index b394feccbb997..7e2bdaf14b3f4 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu-nocontribops-arm64.yml
@@ -6,6 +6,7 @@
 parameters:
   DoEsrp: 'false'
 
+jobs:  
 - template: ../../templates/win-ci-2019.yml
   parameters:
     AgentPool : 'Win-CPU-2019'
@@ -222,6 +223,12 @@ parameters:
       FolderPath: '$(Build.ArtifactStagingDirectory)'
       DoEsrp: ${{ parameters.DoEsrp }}
 
+  - template: ../../templates/validate-nuget.yml
+    parameters:
+      NugetPath: '$(Build.ArtifactStagingDirectory)'
+      PlatformsSupported: 'win-x64,win-x86,win10-arm,linux-x64,osx-x64'
+      VerifyNugetSigning: ${{ parameters.DoEsrp }}
+
   - task: PublishPipelineArtifact@0
     displayName: 'Publish Pipeline NuGet Artifact'
     inputs:
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
index 95672d0be00a0..3e8570bd58ba3 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
@@ -155,6 +155,12 @@ jobs:
       FolderPath: '$(Build.ArtifactStagingDirectory)'
       DoEsrp: ${{ parameters.DoEsrp }}
 
+  - template: ../../templates/validate-nuget.yml
+    parameters:
+      NugetPath: '$(Build.ArtifactStagingDirectory)'
+      PlatformsSupported: 'win-x64,win-x86,linux-x64,osx-x64'
+      VerifyNugetSigning: ${{ parameters.DoEsrp }}
+
   - task: PublishPipelineArtifact@0
     displayName: 'Publish Pipeline NuGet Artifact'
     inputs:
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
index d34ac37afeb04..d2034479cdbf6 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
@@ -112,6 +112,12 @@ jobs:
       FolderPath: '$(Build.ArtifactStagingDirectory)'
       DoEsrp: ${{ parameters.DoEsrp }}
 
+  - template: ../../templates/validate-nuget.yml
+    parameters:
+      NugetPath: '$(Build.ArtifactStagingDirectory)'
+      PlatformsSupported: 'win-x64,linux-x64'
+      VerifyNugetSigning: ${{ parameters.DoEsrp }}
+
   - task: PublishPipelineArtifact@0
     displayName: 'Publish Pipeline NuGet Artifact'
     inputs:
diff --git a/tools/ci_build/github/azure-pipelines/templates/validate-nuget.yml b/tools/ci_build/github/azure-pipelines/templates/validate-nuget.yml
new file mode 100644
index 0000000000000..a162219f5bd9c
--- /dev/null
+++ b/tools/ci_build/github/azure-pipelines/templates/validate-nuget.yml
@@ -0,0 +1,17 @@
+parameters:
+  NugetPath: ''
+  PlatformsSupported: ''
+  VerifyNugetSigning: ''
+
+steps:
+    - task: UsePythonVersion@0
+      displayName: 'Use Python'
+      inputs:
+        versionSpec: 3.7
+
+    - task: PythonScript@0
+      displayName: 'Validate Nuget'
+      inputs:
+        scriptPath: '$(Build.SourcesDirectory)\tools\nuget\validate_nuget.py'
+        arguments: '--nuget_path ${{parameters.NugetPath}} --platforms_supported ${{parameters.PlatformsSupported}} --verify_nuget_signing ${{parameters.VerifyNugetSigning}}'
+        workingDirectory: "$(Build.BinariesDirectory)"
\ No newline at end of file
diff --git a/tools/nuget/validate_nuget.py b/tools/nuget/validate_nuget.py
new file mode 100644
index 0000000000000..744703ac99d94
--- /dev/null
+++ b/tools/nuget/validate_nuget.py
@@ -0,0 +1,140 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+import argparse
+import sys
+import os
+import zipfile # Available Python 3.2 or higher
+import glob
+
+def parse_arguments():
+    parser = argparse.ArgumentParser(description="ONNX Runtime create nuget spec script",
+                                     usage='')
+    # Main arguments
+    parser.add_argument("--nuget_path", required=True, help="Path containing the Nuget to be validated. Must only contain only one Nuget within this.")
+    parser.add_argument("--platforms_supported", required=True, help="Comma separated list (no space). Ex: linux-x64,win-x86,osx-x64")
+    parser.add_argument("--verify_nuget_signing", required=True, help="Flag inidicating if Nuget package signing is to be verified. Only accepets 'true' or 'false'")
+    
+    return parser.parse_args()
+
+def check_exists(path):
+    return os.path.exists(path)
+   
+def is_windows():
+    return sys.platform.startswith("win")
+   
+def check_if_dlls_are_present(platforms_supported, zip_file):
+    platforms = platforms_supported.strip().split(",")
+    for platform in platforms:
+        if platform.startswith("win"):
+            path = "runtimes/" + platform + "/native/onnxruntime.dll"
+            print('Checking path: ' + path)
+            if (not path in zip_file.namelist()):
+               print("onnxruntime.dll not found for " + platform)
+               print(zip_file.namelist())
+               raise Exception("onnxruntime.dll not found for " + platform)
+
+        elif platform.startswith("linux"):
+            path = "runtimes/" + platform + "/native/libonnxruntime.so"
+            print('Checking path: ' + path)
+            if (not path in zip_file.namelist()):
+               print("libonnxruntime.so not found for " + platform)
+               raise Exception("libonnxruntime.so not found for " + platform)
+               
+        elif platform.startswith("osx"):
+            path = "runtimes/" + platform + "/native/libonnxruntime.dylib"
+            print('Checking path: ' + path)        
+            if (not path in zip_file.namelist()):
+               print("libonnxruntime.dylib not found for " + platform)
+               raise Exception("libonnxruntime.dylib not found for " + platform)
+               
+        else:
+            raise Exception("Unsupported platform: " + platform)
+            
+def check_if_nuget_is_signed(nuget_path):
+    code_sign_summary_file = glob.glob(os.path.join(nuget_path, '*.md'))
+    if (len(code_sign_summary_file) != 1):
+        print('CodeSignSummary files found in path: ')
+        print(code_sign_summary_file)
+        raise Exception('No CodeSignSummary files / more than one CodeSignSummary files found in the given path.')
+     
+    print('CodeSignSummary file: ' + code_sign_summary_file[0])
+    
+    with open(code_sign_summary_file[0]) as f:
+        contents = f.read()
+        return "Pass" in contents
+
+    return False    
+            
+def main():   
+    args = parse_arguments()
+
+    files = os.listdir(args.nuget_path)
+    nuget_packages_found_in_path = [i for i in files if i.endswith('.nupkg')]
+    if (len(nuget_packages_found_in_path) != 1):
+        print('Nuget packages found in path: ')
+        print(nuget_packages_found_in_path)
+        raise Exception('No Nuget packages / more than one Nuget packages found in the given path.')
+    
+    nuget_file_name = nuget_packages_found_in_path[0]
+    full_nuget_path = os.path.join(args.nuget_path, nuget_file_name)
+    
+    exit_code = 0
+            
+    nupkg_copy_name = "NugetCopy" + ".nupkg" 
+    zip_copy_name = "NugetCopy" + ".zip"
+    zip_file = None
+
+    # Remove any residual files
+    if check_exists(nupkg_copy_name):
+        os.remove(nupkg_copy_name)
+        
+    if check_exists(zip_copy_name):
+        os.remove(zip_copy_name)
+         
+    # Do all validations here         
+    try:
+        if not is_windows():
+            raise Exception('Nuget validation is currently supported only on Windows')
+            
+        # Make a copy of the Nuget package
+        print('Making a copy of the Nuget and extracting its contents') 
+        os.system("copy "  + full_nuget_path  + " " + nupkg_copy_name)
+        
+        # Convert nupkg to zip
+        os.rename(nupkg_copy_name, zip_copy_name)
+        zip_file = zipfile.ZipFile(zip_copy_name)
+        
+        # Check if the relevant dlls are present in the Nuget/Zip
+        print('Checking if the Nuget contains relevant dlls')
+        check_if_dlls_are_present(args.platforms_supported, zip_file)
+
+        # Check if the Nuget has been signed
+        if (args.verify_nuget_signing != 'true' and args.verify_nuget_signing != 'false'):
+            raise Exception('Parameter verify_nuget_signing accepts only true or false as an argument');
+            
+        if (args.verify_nuget_signing == 'true'):
+            print('Verifying if Nuget has been signed')
+            if(not check_if_nuget_is_signed(args.nuget_path)):
+                print('Nuget signing verification failed')
+                raise Exception('Nuget signing verification failed')
+        
+    except:
+        exit_code = 1
+        
+    finally:
+        print('Cleaning up after Nuget validation')
+            
+        if zip_file is not None:
+            zip_file.close()
+
+        if check_exists(zip_copy_name):
+            os.remove(zip_copy_name)
+
+        if exit_code == 0:
+            print('Nuget validation was successful')
+        else:
+            raise Exception('Nuget validation was unsuccessful')
+        
+if __name__ == "__main__":
+    sys.exit(main())

From f84240db2bfe314fc325269800d4ca0a6df3d5c0 Mon Sep 17 00:00:00 2001
From: Nathan <7902510+ybrnathan@users.noreply.github.com>
Date: Wed, 8 Jan 2020 09:59:42 -0800
Subject: [PATCH 33/75] add uint8 support to where op (#2792)

---
 onnxruntime/core/providers/cpu/cpu_execution_provider.cc   | 3 +++
 onnxruntime/core/providers/cpu/tensor/where_op.cc          | 2 +-
 onnxruntime/core/providers/cuda/cuda_execution_provider.cc | 2 ++
 onnxruntime/core/providers/cuda/tensor/where.cc            | 1 +
 onnxruntime/core/providers/cuda/tensor/where_impl.cu       | 1 +
 5 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
index 625b6a9c40a05..496d96bba76dd 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
@@ -270,6 +270,7 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain,
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, float, Where);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, int32_t, Where);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, int64_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, uint8_t, Where);
 class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10, Flatten);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, BatchNormalization);
 class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10, Gemm);
@@ -890,6 +891,8 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
                                                                   Where)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, int64_t,
                                                                   Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, uint8_t,
+                                                                  Where)>,                                                                  
       BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
                                                                       Flatten)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9,
diff --git a/onnxruntime/core/providers/cpu/tensor/where_op.cc b/onnxruntime/core/providers/cpu/tensor/where_op.cc
index d518352af85a0..7585943c00453 100644
--- a/onnxruntime/core/providers/cpu/tensor/where_op.cc
+++ b/onnxruntime/core/providers/cpu/tensor/where_op.cc
@@ -22,7 +22,7 @@ namespace onnxruntime {
   WHERE_TYPED_KERNEL_WITH_TYPE_NAME(type, type)
 
 // start with a subset of types, enable more as needed...
-//WHERE_TYPED_KERNEL(uint8_t)
+WHERE_TYPED_KERNEL(uint8_t)
 //WHERE_TYPED_KERNEL(uint16_t)
 //WHERE_TYPED_KERNEL(uint32_t)
 //WHERE_TYPED_KERNEL(uint64_t)
diff --git a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
index 375d714612c1f..92f7ba87e6b70 100644
--- a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
+++ b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
@@ -539,6 +539,7 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain,
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, Where);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, Where);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int64_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, Where);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, bool, NonZero);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, NonZero);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, NonZero);
@@ -1004,6 +1005,7 @@ static void RegisterCudaKernels(KernelRegistry& kernel_registry) {
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, Where)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, Where)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int64_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, Where)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, bool, NonZero)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, NonZero)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, NonZero)>,
diff --git a/onnxruntime/core/providers/cuda/tensor/where.cc b/onnxruntime/core/providers/cuda/tensor/where.cc
index 142dc9294414e..1e341f3f072cb 100644
--- a/onnxruntime/core/providers/cuda/tensor/where.cc
+++ b/onnxruntime/core/providers/cuda/tensor/where.cc
@@ -183,6 +183,7 @@ Status Where<T>::ComputeInternal(OpKernelContext* context) const {
 #define SPECIALIZED_COMPUTE(T) \
   SPECIALIZED_COMPUTE_WITH_NAME(T, T)
 
+SPECIALIZED_COMPUTE(uint8_t)
 SPECIALIZED_COMPUTE(int32_t)
 SPECIALIZED_COMPUTE(int64_t)
 SPECIALIZED_COMPUTE(float)
diff --git a/onnxruntime/core/providers/cuda/tensor/where_impl.cu b/onnxruntime/core/providers/cuda/tensor/where_impl.cu
index 7bde79a2765da..6c7f92dcf7197 100644
--- a/onnxruntime/core/providers/cuda/tensor/where_impl.cu
+++ b/onnxruntime/core/providers/cuda/tensor/where_impl.cu
@@ -120,6 +120,7 @@ void WhereImpl(
                              T* output_data,                         \
                              size_t count);
 
+SPECIALIZED_IMPL(uint8_t)
 SPECIALIZED_IMPL(int32_t)
 SPECIALIZED_IMPL(int64_t)
 SPECIALIZED_IMPL(float)

From 32c5e76a16eba802c9faa689f275fd792905649d Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Wed, 8 Jan 2020 11:32:27 -0800
Subject: [PATCH 34/75] Improve bert optimization script: (#2712)

(1) Move input int64=>int32 conversion to embed layer fusion.
(2) Output epsilon attribute for LayerNormalization fusion.
---
 .../tools/bert/bert_model_optimization.py     | 140 +++++++++++++-----
 1 file changed, 101 insertions(+), 39 deletions(-)

diff --git a/onnxruntime/python/tools/bert/bert_model_optimization.py b/onnxruntime/python/tools/bert/bert_model_optimization.py
index 4f074a8433a7c..2508f7f64b8bc 100644
--- a/onnxruntime/python/tools/bert/bert_model_optimization.py
+++ b/onnxruntime/python/tools/bert/bert_model_optimization.py
@@ -6,6 +6,18 @@
 # Convert Bert ONNX model exported from PyTorch to use Attention, Gelu,
 # SkipLayerNormalization and EmbedLayerNormalization ops to optimize
 # performance on NVidia GPU.
+
+# Note: This script is not required for Bert model optimization. 
+# OnnxRuntime has bert model optimization support internally. The recommended way is
+# to set optimization level to ORT_ENABLE_EXTENDED during Bert model inference.
+# See the following document for more information:
+# https://github.com/microsoft/onnxruntime/blob/master/docs/ONNX_Runtime_Graph_Optimizations.md
+
+# This script is retained for experiment purpose. Useful senarios like the following:
+#  (1) Change model from fp32 to fp16.
+#  (2) Change input data type from int64 to int32.
+#  (3) Model cannot be handled to OnnxRuntime graph optimization, and you can modify this script to get optimized model.
+
 import onnx
 import sys
 import argparse
@@ -193,6 +205,13 @@ def get_constant_value(self, output_name):
                 for att in node.attribute:
                     if att.name == 'value':
                         return numpy_helper.to_array(att.t)
+
+        # Fall back to intializer since constant folding might have been
+        # applied.
+        initializer = self.get_initializer(output_name)
+        if initializer is not None:
+            return numpy_helper.to_array(initializer)
+
         return None
 
     def get_constant_input(self, node):
@@ -200,13 +219,14 @@ def get_constant_input(self, node):
             value = self.get_constant_value(input)
             if value is not None:
                 return i, value
+
         return None, None
 
     def find_constant_input(self, node, expected_value, delta=0.000001):
-        for i, input in enumerate(node.input):
-            value = self.get_constant_value(input)
-            if value is not None and value.size == 1 and abs(value - expected_value) < delta:
-                return i
+        i, value = self.get_constant_input(node)
+        if value is not None and value.size == 1 and abs(value - expected_value) < delta:
+            return i
+
         return -1
 
     def has_constant_input(self, node, expected_value, delta=0.000001):
@@ -402,6 +422,9 @@ def __init__(self, model, num_heads, hidden_size, sequence_length, verbose):
 
         # A lookup table with mask input as key, and mask index output as value
         self.mask_indice = {}
+        # A lookup table with mask input as key, and cast (to int32) output as value
+        self.mask_casted = {}
+
         self.bert_inputs = []
 
         # constant node names
@@ -414,21 +437,52 @@ def get_normalize_nodes(self):
     def normalize_children_types(self):
         return ['MatMul', 'MatMul', 'MatMul', 'SkipLayerNormalization']
 
+    def cast_graph_input_to_int32(self, input_name):
+        graph_input = self.find_graph_input(input_name)
+        if graph_input is not None and graph_input.type.tensor_type.elem_type != TensorProto.INT32:
+            cast_output = input_name + '_int32'
+            cast_node = onnx.helper.make_node('Cast', inputs=[input_name], outputs=[cast_output])
+            cast_node.attribute.extend([onnx.helper.make_attribute("to", int(TensorProto.INT32))])
+            self.add_node(cast_node)
+            return True, cast_output
+
+        return False, input_name
+
+    def undo_cast_input_to_int32(self, input_name):
+        input_name_to_nodes = self.input_name_to_nodes()
+        nodes =  input_name_to_nodes[input_name]
+        for node in nodes:
+            if node.op_type == "Cast":
+                is_int32 = False
+                for att in node.attribute:
+                    if att.name == 'to' and att.i == int(TensorProto.INT32):
+                        is_int32 = True
+                        break
+                if is_int32:
+                    output_name = node.output[0]
+                    self.remove_node(node)
+                    self.replace_input_of_all_nodes(output_name, input_name)
+
     def process_mask(self, input):
         if input in self.mask_indice:
             return self.mask_indice[input]
 
+        # Add cast to convert int64 to int32
+        casted, input_name = self.cast_graph_input_to_int32(input)
+        if casted:
+            self.mask_casted[input] = input_name
+
         # Add a mask processing node
         output_name = self.create_node_name('mask_index')
         mask_index_node = onnx.helper.make_node('ReduceSum',
-            inputs=[input],
+            inputs=[input_name],
             outputs=[output_name],
             name=self.create_node_name('ReduceSum', 'MaskReduceSum'))
         mask_index_node.attribute.extend([onnx.helper.make_attribute("axes", [1]), onnx.helper.make_attribute("keepdims", 0)])
         self.add_node(mask_index_node)
+        
         self.mask_indice[input] = output_name
-
-        return self.mask_indice[input]
+        return output_name
 
     def create_attention_node(self, mask_index, q_matmul, k_matmul, v_matmul, q_add, k_add, v_add, input, output):
         q_weight = self.get_initializer(q_matmul.input[1])
@@ -437,7 +491,7 @@ def create_attention_node(self, mask_index, q_matmul, k_matmul, v_matmul, q_add,
         q_bias = self.get_initializer(q_add.input[1])
         k_bias = self.get_initializer(k_add.input[1])
         v_bias = self.get_initializer(v_add.input[1])
-
+        
         qw = numpy_helper.to_array(q_weight)
         assert qw.shape == (self.hidden_size, self.hidden_size)
 
@@ -932,9 +986,9 @@ def fuse_reshape(self):
         |                |                   v            v
         +---(optional graph)               SkipLayerNormalization
 
-      Optional graph is used to generate position list (0, 1, ...). It can be a constant in some model.
+      Optional graph is used to generate position list (0, 1, ...) per batch. It can be a constant in some model.
     """
-    def fuse_embed_layer(self):
+    def fuse_embed_layer(self, input_int32):
         nodes = self.nodes()
         input_name_to_nodes = self.input_name_to_nodes()
         output_name_to_node = self.output_name_to_node()
@@ -972,9 +1026,13 @@ def fuse_embed_layer(self):
 
         position_embedding_path = self.match_parent_path(add_node, ['Gather', 'Expand', 'Shape'], [1, 1, 1])
         if position_embedding_path is None:
-            print("Failed to find position embedding")
-            return
-        position_embedding_gather, position_embedding_expand, position_embedding_shape = position_embedding_path
+            position_embedding_path2 = self.match_parent_path(add_node, ['Gather', 'Expand', 'Concat', 'Unsqueeze', 'Gather', 'Shape'], [1, 1, 1, 1, 0, 0])
+            if position_embedding_path2 is None:
+                print("Failed to find position embedding")
+                return
+            position_embedding_gather, position_embedding_expand, _, _, _, position_embedding_shape = position_embedding_path2
+        else:
+            position_embedding_gather, position_embedding_expand, position_embedding_shape = position_embedding_path
 
         segment_embedding_path = self.match_parent_path(normalize_node, ['Gather'], [1])
         if segment_embedding_path is None:
@@ -995,19 +1053,32 @@ def fuse_embed_layer(self):
         nodes_to_remove.extend([normalize_node, add_node, segment_embedding_gather, word_embedding_gather, position_embedding_gather, position_embedding_expand])
         nodes_to_remove.extend([mask_node])
 
+        # store inputs for further processing
+        self.bert_inputs = [input_ids, segment_ids, mask_input_name]
+
+        if not input_int32:
+            # When mask has been casted to int32, use that casted one as input of embed layer norm.
+            if mask_input_name in self.mask_casted:
+                mask_input_name = self.mask_casted[mask_input_name]
+
+            # Cast input_ids and segment_ids to int32.
+            casted, input_ids = self.cast_graph_input_to_int32(input_ids)
+
+            casted, segment_ids = self.cast_graph_input_to_int32(segment_ids)
+        else:
+            self.undo_cast_input_to_int32(mask_input_name)
+
         embed_node = onnx.helper.make_node('EmbedLayerNormalization',
-                        inputs=[input_ids, segment_ids, 
+                        inputs=[input_ids,
+                                segment_ids, 
                                 word_embedding_gather.input[0], position_embedding_gather.input[0], segment_embedding_gather.input[0],
                                 normalize_node.input[2], normalize_node.input[3], # gamma and beta
                                 mask_input_name],
-                        outputs=["embed_output", self.mask_indice[mask_input_name]],
+                        outputs=["embed_output", mask_output_name],
                         name="EmbedLayer")
 
         embed_node.domain = "com.microsoft"
         
-        # store inputs for further processing
-        self.bert_inputs = [input_ids, segment_ids, mask_input_name]
-
         self.replace_input_of_all_nodes(normalize_node.output[0], 'embed_output')
 
         self.remove_nodes(nodes_to_remove)
@@ -1015,8 +1086,12 @@ def fuse_embed_layer(self):
         self.update_graph()
         print("Fused EmbedLayerNormalization count: 1")
 
-    def get_bert_inputs(self):
-        return self.bert_inputs
+        # Change graph input data type int32 if needed.
+        if input_int32:
+            self.change_input_to_int32()
+
+    def get_bert_inputs(self, include_mask=True):
+        return self.bert_inputs if include_mask else self.bert_inputs[:2]
 
     def get_batch_size_from_graph_input(self):
         graph = self.graph()
@@ -1040,6 +1115,7 @@ def change_input_to_int32(self):
         batch_size = self.get_batch_size_from_graph_input()
         input_batch_size = batch_size if isinstance(batch_size, int) else 1
         new_graph_inputs = []
+
         bert_inputs = self.get_bert_inputs()
         for input in graph.input:
             if input.name in bert_inputs:
@@ -1063,16 +1139,6 @@ def change_input_to_int32(self):
         # restore opset version
         self.model.opset_import[0].version = original_opset_version
 
-    def cast_input_to_int32(self):
-        bert_inputs = self.get_bert_inputs()
-        for input in bert_inputs:
-            graph_input = self.find_graph_input(input)
-            if graph_input is not None and graph_input.type.tensor_type.elem_type != TensorProto.INT32:
-                cast_output = input + '_int32'
-                cast_node = onnx.helper.make_node('Cast', inputs=[input], outputs=[cast_output])
-                cast_node.attribute.extend([onnx.helper.make_attribute("to", int(TensorProto.INT32))])
-                self.replace_input_of_all_nodes(input, cast_output)
-                self.add_node(cast_node)
 
     # Update input and output using dynamic batch
     def update_dynamic_batch_io(self, dynamic_batch_dim='batch'):
@@ -1095,7 +1161,7 @@ def update_dynamic_batch_io(self, dynamic_batch_dim='batch'):
           |                      |
           |                      v
         Add --> ReduceMean -->  Sub  --> Pow --> ReduceMean --> Add --> Sqrt --> Div --> Mul --> Add
-                 (axis=2 or -1)  |      (Y=2)   (axis=2 or -1)  (E-6 or E-12)    ^
+                 (axis=2 or -1)  |      (Y=2)   (axis=2 or -1)  (E-6 or E-12 or 0)    ^
                                  |                                               |
                                  +-----------------------------------------------+
 
@@ -1200,6 +1266,7 @@ def fuse_layer_norm(self):
                     normalize_node = onnx.helper.make_node('LayerNormalization',
                         inputs=[node.input[0], weight_input, bias_input],
                         outputs=[last_add_node.output[0]])
+                    normalize_node.attribute.extend([onnx.helper.make_attribute("epsilon", add_weight)])
                     layernorm_nodes.extend([normalize_node])
 
         self.remove_nodes(nodes_to_remove)
@@ -1250,7 +1317,7 @@ def main():
 
     bert_model.fuse_attention()
 
-    bert_model.fuse_embed_layer()
+    bert_model.fuse_embed_layer(args.input_int32)
 
     # Fuse Gelu and Add Bias before it.
     bert_model.fuse_add_bias_gelu()
@@ -1258,12 +1325,6 @@ def main():
     # Fuse SkipLayerNormalization and Add Bias before it.
     bert_model.fuse_add_bias_skip_layer_norm()
 
-    if bert_model.get_bert_inputs():
-        if args.input_int32:
-            bert_model.change_input_to_int32()
-        else:
-            bert_model.cast_input_to_int32()
-
     if args.float16:
         bert_model.convert_model_float32_to_float16()
 
@@ -1277,4 +1338,5 @@ def main():
     with open(args.output, "wb") as out:
         out.write(bert_model.model.SerializeToString())
 
-main()
+if __name__ == "__main__":
+    main()

From 1978376e1eff681d02bf69c10d8af2821b821295 Mon Sep 17 00:00:00 2001
From: George Wu <George.Wu@microsoft.com>
Date: Wed, 8 Jan 2020 11:17:48 -1000
Subject: [PATCH 35/75] add session creation time cost. (#2798)

---
 .../test/perftest/performance_runner.cc       | 19 ++++++++++++-------
 .../test/perftest/performance_runner.h        |  2 ++
 2 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/onnxruntime/test/perftest/performance_runner.cc b/onnxruntime/test/perftest/performance_runner.cc
index 7820639946fad..6e539162bf88a 100644
--- a/onnxruntime/test/perftest/performance_runner.cc
+++ b/onnxruntime/test/perftest/performance_runner.cc
@@ -69,15 +69,17 @@ Status PerformanceRunner::Run() {
   performance_result_.average_CPU_usage = p_ICPUUsage->GetUsage();
   performance_result_.peak_workingset_size = utils::GetPeakWorkingSetSize();
 
+  std::chrono::duration<double> session_create_duration = session_create_end_ - session_create_start_;
   // TODO: end profiling
   // if (!performance_test_config_.run_config.profile_file.empty()) session_object->EndProfiling();
-  std::chrono::duration<double> duration = performance_result_.end_ - performance_result_.start_;
+  std::chrono::duration<double> inference_duration = performance_result_.end_ - performance_result_.start_;
 
-  std::cout << "Total time cost:" << performance_result_.total_time_cost << std::endl  // sum of time taken by each request
-            << "Total iterations:" << performance_result_.time_costs.size() << std::endl
-            << "Average time cost:" << performance_result_.total_time_cost / performance_result_.time_costs.size() * 1000 << " ms" << std::endl
+  std::cout << "Session creation time cost:" << session_create_duration.count() << " s" << std::endl
+            << "Total inference time cost:" << performance_result_.total_time_cost << " s" << std::endl  // sum of time taken by each request
+            << "Total inference requests:" << performance_result_.time_costs.size() << std::endl
+            << "Average inference time cost:" << performance_result_.total_time_cost / performance_result_.time_costs.size() * 1000 << " ms" << std::endl
             // Time between start and end of run. Less than Total time cost when running requests in parallel.
-            << "Total run time:" << duration.count() << " s" << std::endl;
+            << "Total inference run time:" << inference_duration.count() << " s" << std::endl;
   return Status::OK();
 }
 
@@ -191,8 +193,11 @@ static TestSession* CreateSession(Ort::Env& env, std::random_device& rd,
 }
 PerformanceRunner::PerformanceRunner(Ort::Env& env, const PerformanceTestConfig& test_config, std::random_device& rd)
     : performance_test_config_(test_config),
-      test_model_info_(CreateModelInfo(test_config)),
-      session_(CreateSession(env, rd, test_config, test_model_info_)) {}
+      test_model_info_(CreateModelInfo(test_config)) {
+  session_create_start_ = std::chrono::high_resolution_clock::now();
+  session_.reset(CreateSession(env, rd, test_config, test_model_info_));
+  session_create_end_ = std::chrono::high_resolution_clock::now();
+}
 
 PerformanceRunner::~PerformanceRunner() = default;
 
diff --git a/onnxruntime/test/perftest/performance_runner.h b/onnxruntime/test/perftest/performance_runner.h
index bd3d03e80475f..902422bb4af66 100644
--- a/onnxruntime/test/perftest/performance_runner.h
+++ b/onnxruntime/test/perftest/performance_runner.h
@@ -132,6 +132,8 @@ class PerformanceRunner {
   }
 
  private:
+  std::chrono::time_point<std::chrono::high_resolution_clock> session_create_start_;
+  std::chrono::time_point<std::chrono::high_resolution_clock> session_create_end_;
   PerformanceResult performance_result_;
   PerformanceTestConfig performance_test_config_;
   TestModelInfo* test_model_info_;

From 2c8179bee44a53cc51ecd1c78adf43dcc2f19d3b Mon Sep 17 00:00:00 2001
From: Dmitri Smirnov <yuslepukhin@users.noreply.github.com>
Date: Thu, 9 Jan 2020 10:54:12 -0800
Subject: [PATCH 36/75] ML.NET team needs featurizers within a package (#2789)

Add auto ml featurizers to Windows, MacOS as well as to GPU  packaging-pipelines.
---
 .../github/azure-pipelines/nuget/templates/cpu.yml        | 8 ++++----
 .../github/azure-pipelines/nuget/templates/gpu.yml        | 4 ++--
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
index 3e8570bd58ba3..b1239966d645c 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/cpu.yml
@@ -14,7 +14,7 @@ jobs:
     AgentPool : 'Win-CPU-2019'
     ArtifactName: 'drop-nuget'
     JobName: 'Windows_CI_Dev'
-    BuildCommand:  '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_openmp --build_shared_lib --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019"'
+    BuildCommand:  '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_openmp --build_shared_lib --use_featurizers --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019"'
     BuildArch: 'x64'
     msbuildArchitecture: 'amd64'
     EnvSetupScript: 'setup_env.bat'
@@ -34,7 +34,7 @@ jobs:
     AgentPool : 'Win-CPU-2019'
     ArtifactName: 'drop-win-x86-zip'
     JobName: 'Windows_CI_Dev_x86'
-    BuildCommand: '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_openmp --build_shared_lib --enable_onnx_tests --x86 --use_telemetry --cmake_generator "Visual Studio 16 2019"'
+    BuildCommand: '--build_dir $(Build.BinariesDirectory) --skip_submodule_sync --use_openmp --build_shared_lib --use_featurizers --enable_onnx_tests --x86 --use_telemetry --cmake_generator "Visual Studio 16 2019"'
     BuildArch: 'x86'
     msbuildArchitecture: 'x86'
     EnvSetupScript: 'setup_env_x86.bat'
@@ -66,7 +66,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-centos6 /bin/bash -c "/usr/bin/python3.6 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --enable_onnx_tests && cd /build/Release && make install DESTDIR=/build/linux-x64"
+          docker run --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-centos6 /bin/bash -c "/usr/bin/python3.6 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_featurizers --use_openmp --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --enable_onnx_tests && cd /build/Release && make install DESTDIR=/build/linux-x64"
         workingDirectory: $(Build.SourcesDirectory)
     - script: |
        set -e -x
@@ -92,7 +92,7 @@ jobs:
   parameters:
     AgentPool : $(AgentPoolMacOS)
     JobName: 'MacOS_CI_Dev'
-    BuildCommand: 'python3 $(Build.SourcesDirectory)/tools/ci_build/build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --parallel --build_shared_lib --use_openmp --enable_onnx_tests --config RelWithDebInfo'
+    BuildCommand: 'python3 $(Build.SourcesDirectory)/tools/ci_build/build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --parallel --build_shared_lib --use_featurizers --use_openmp --enable_onnx_tests --config RelWithDebInfo'
     DoNugetPack : 'true'
     NuPackScript: |
      set -e -x
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
index d2034479cdbf6..239a98d2d151d 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
@@ -13,7 +13,7 @@ jobs:
     AgentPool : 'Win-GPU-2019'
     ArtifactName: 'drop-nuget'    
     JobName: 'Windows_CI_GPU_Dev'
-    BuildCommand: --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019" --msvc_toolset 14.16 --use_cuda --cuda_version=10.0 --cuda_home="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"
+    BuildCommand: --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --use_featurizers --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019" --msvc_toolset 14.16 --use_cuda --cuda_version=10.0 --cuda_home="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"
     BuildArch: 'x64'
     msbuildArchitecture: 'amd64'
     EnvSetupScript: 'setup_env.bat'    
@@ -43,7 +43,7 @@ jobs:
     - task: CmdLine@2
       inputs:
         script: |
-          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-centos6-gpu /bin/bash -c "/usr/bin/python3.6 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_featurizers --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0 --enable_onnx_tests && cd /build/Release && make install DESTDIR=/build/linux-x64"
+          docker run --gpus all -e NVIDIA_VISIBLE_DEVICES=all --rm --volume $(Build.SourcesDirectory):/onnxruntime_src --volume $(Build.BinariesDirectory):/build -e NIGHTLY_BUILD onnxruntime-centos6-gpu /bin/bash -c "/usr/bin/python3.6 /onnxruntime_src/tools/ci_build/build.py --build_dir /build --config Release --skip_submodule_sync  --parallel --build_shared_lib --use_featurizers --cmake_path /usr/bin/cmake --ctest_path /usr/bin/ctest --use_cuda --cuda_version=10.0 --cuda_home=/usr/local/cuda-10.0  --cudnn_home=/usr/local/cuda-10.0 --enable_onnx_tests && cd /build/Release && make install DESTDIR=/build/linux-x64"
     - script: |
        set -e -x
        mv $(Build.BinariesDirectory)/linux-x64/usr/local/lib64 $(Build.BinariesDirectory)/linux-x64/linux-x64

From 71b5165ed39d13549b9a83194c283e2fc898b2a5 Mon Sep 17 00:00:00 2001
From: Yufeng Li <liyufeng1987@gmail.com>
Date: Thu, 9 Jan 2020 13:48:18 -0800
Subject: [PATCH 37/75] Initialize max of softmax with lowest of float (#2786)

---
 onnxruntime/contrib_ops/cpu/bert/attention.cc |  3 +-
 .../contrib_ops/cuda/bert/attention_impl.cu   | 43 ++++++++++---------
 2 files changed, 23 insertions(+), 23 deletions(-)

diff --git a/onnxruntime/contrib_ops/cpu/bert/attention.cc b/onnxruntime/contrib_ops/cpu/bert/attention.cc
index 9ec7e849c0cb6..ccef01924b3ef 100644
--- a/onnxruntime/contrib_ops/cpu/bert/attention.cc
+++ b/onnxruntime/contrib_ops/cpu/bert/attention.cc
@@ -242,8 +242,7 @@ Status Attention<T>::Compute(OpKernelContext* context) const {
       // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
       // a math transform as below is leveraged to get a stable softmax:
       // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
-      // And for convenience, force max to 0.f if all xi are negative
-      float max = 0.f;
+      float max = -std::numeric_limits<float>::infinity();
       for (int i = 0; i < D; i++) {
         if (max < x[i]) max = x[i];
       }
diff --git a/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu b/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu
index 5c72a1d53e316..26aee9affb46f 100644
--- a/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu
+++ b/onnxruntime/contrib_ops/cuda/bert/attention_impl.cu
@@ -24,6 +24,7 @@ limitations under the License.
 #include <cub/cub.cuh>
 #include <cublas_v2.h>
 #include <cuda_fp16.h>
+#include <math_constants.h>
 #include "core/providers/cuda/cu_inc/common.cuh"
 #include "core/providers/cuda/cuda_common.h"
 #include "attention_impl.h"
@@ -61,22 +62,21 @@ __device__ inline void Softmax(const int ld, const int num_valid, const T* input
   __shared__ float sum_reverse_block;
   __shared__ float max_block;
 
-  float thread_data(0);
+  float thread_data_max(-CUDART_INF_F);
 
+  // e^x is represented as infinity if x is large enough, like 100.f.
+  // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
+  // a math transform as below is leveraged to get a stable softmax:
+  // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
   const int offset = (blockIdx.y * gridDim.x + blockIdx.x) * ld;
   for (int i = threadIdx.x; i < num_valid; i += TPB) {
     const int index = offset + i;
-    if (thread_data < float(input[index])) {
-      thread_data = float(input[index]);
+    if (thread_data_max < float(input[index])) {
+      thread_data_max = float(input[index]);
     }
   }
 
-  // e^x is represented as infinity if x is large enough, like 100.f.
-  // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
-  // a math transform as below is leveraged to get a stable softmax:
-  // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
-  // And for convenience, force max to 0.f if all xi are negative
-  const auto max = BlockReduce(tmp_storage).Reduce(thread_data, cub::Max());
+  const auto max = BlockReduce(tmp_storage).Reduce(thread_data_max, cub::Max());
 
   // Store max value
   if (threadIdx.x == 0) {
@@ -84,13 +84,14 @@ __device__ inline void Softmax(const int ld, const int num_valid, const T* input
   }
   __syncthreads();
 
+  float thread_data_sum(0.f);
   for (int i = threadIdx.x; i < num_valid; i += TPB) {
     const int index = offset + i;
     const float val = input[index];
-    thread_data += expf(val - max_block);
+    thread_data_sum += expf(val - max_block);
   }
 
-  const auto sum = BlockReduce(tmp_storage).Reduce(thread_data, cub::Sum());
+  const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_sum, cub::Sum());
   if (threadIdx.x == 0) {
     sum_reverse_block = 1.f / sum;
   }
@@ -114,17 +115,16 @@ __device__ inline void SoftmaxSmall(const int ld, const int num_valid, const T*
   const int offset = (blockIdx.y * gridDim.x + blockIdx.x) * ld;
   const int index = offset + threadIdx.x;
 
-  float thread_data(0);
-  if (threadIdx.x < num_valid) {
-    thread_data = input[index];
-  }
-
   // e^x is represented as infinity if x is large enough, like 100.f.
   // Infinity divided by Infinity is a NAN. Thus, softmax gets a NAN if one or more item are large enough.
   // a math transform as below is leveraged to get a stable softmax:
   // e^xi/(e^x1 + ...e^xn) = e^(xi - max) / (e^(x1 - max) + ... + e^(xn - max))
-  // And for convenience, force max to 0.f if all xi are negative
-  const auto max = BlockReduce(tmp_storage).Reduce(thread_data, cub::Max(), num_valid);
+  float thread_data_max(-CUDART_INF_F);
+  if (threadIdx.x < num_valid) {
+    thread_data_max = input[index];
+  }
+
+  const auto max = BlockReduce(tmp_storage).Reduce(thread_data_max, cub::Max(), num_valid);
 
   // Store max value
   if (threadIdx.x == 0) {
@@ -132,12 +132,13 @@ __device__ inline void SoftmaxSmall(const int ld, const int num_valid, const T*
   }
   __syncthreads();
 
+  float thread_data_exp(0.f);
   if (threadIdx.x < num_valid) {
     const float val = input[index];
-    thread_data = expf(val - max_block);
+    thread_data_exp = expf(val - max_block);
   }
 
-  const auto sum = BlockReduce(tmp_storage).Reduce(thread_data, cub::Sum(), num_valid);
+  const auto sum = BlockReduce(tmp_storage).Reduce(thread_data_exp, cub::Sum(), num_valid);
 
   // Store max value
   if (threadIdx.x == 0) {
@@ -147,7 +148,7 @@ __device__ inline void SoftmaxSmall(const int ld, const int num_valid, const T*
 
   if (threadIdx.x < ld) {
     // this will be 0 for threadIdx.x >= num_valid
-    output[index] = T(thread_data * sum_reverse_block);
+    output[index] = T(thread_data_exp * sum_reverse_block);
   }
 }
 

From 7ef6570e2715569dc907208855e702c7b8076484 Mon Sep 17 00:00:00 2001
From: Tracy Sharpe <42477615+tracysh@users.noreply.github.com>
Date: Thu, 9 Jan 2020 14:48:20 -0800
Subject: [PATCH 38/75] MLAS: update SGEMM threading parameters (#2808)

---
 onnxruntime/core/mlas/lib/mlasi.h | 13 -------------
 1 file changed, 13 deletions(-)

diff --git a/onnxruntime/core/mlas/lib/mlasi.h b/onnxruntime/core/mlas/lib/mlasi.h
index 4ce094eb2fee6..ff0b038cf200a 100644
--- a/onnxruntime/core/mlas/lib/mlasi.h
+++ b/onnxruntime/core/mlas/lib/mlasi.h
@@ -554,21 +554,8 @@ extern "C" {
 // Define the target number of per-thread multiplies before using another
 // thread to perform additional work.
 //
-// The number is derived from performance results running SGEMM across a
-// range of workloads and observing the ideal number of threads to complete
-// that workload.
-//
 
-#if defined(_OPENMP)
 #define MLAS_SGEMM_THREAD_COMPLEXITY                (64 * 1024)
-#else
-#if defined(MLAS_TARGET_AMD64)
-#define MLAS_SGEMM_THREAD_COMPLEXITY                (2 * 1024 * 1024)
-#else
-#define MLAS_SGEMM_THREAD_COMPLEXITY                (1 * 1024 * 1024)
-#endif
-#endif
-
 #define MLAS_DGEMM_THREAD_COMPLEXITY                (64 * 1024)
 
 //

From b0019ac7fef58c40d731850c74403fe706e85c77 Mon Sep 17 00:00:00 2001
From: Ke Zhang <kezhan@microsoft.com>
Date: Thu, 9 Jan 2020 16:52:34 -0800
Subject: [PATCH 39/75] add interface to copy batch tensors. (#2807)

* add interface to copy batch tensors.

* onnxruntime
---
 include/onnxruntime/core/session/onnxruntime_c_api.h | 12 ++++++------
 onnxruntime/core/framework/data_transfer.cc          |  8 ++++++++
 onnxruntime/core/framework/data_transfer.h           |  1 +
 onnxruntime/core/session/onnxruntime_c_api.cc        | 10 +---------
 4 files changed, 16 insertions(+), 15 deletions(-)

diff --git a/include/onnxruntime/core/session/onnxruntime_c_api.h b/include/onnxruntime/core/session/onnxruntime_c_api.h
index c5a42d66cfcde..ed07c3bbb2d87 100644
--- a/include/onnxruntime/core/session/onnxruntime_c_api.h
+++ b/include/onnxruntime/core/session/onnxruntime_c_api.h
@@ -253,7 +253,7 @@ struct OrtApi {
 
   // TODO: document the path separator convention? '/' vs '\'
   // TODO: should specify the access characteristics of model_path. Is this read only during the
-  // execution of OrtCreateSession, or does the OrtSession retain a handle to the file/directory
+  // execution of CreateSession, or does the OrtSession retain a handle to the file/directory
   // and continue to access throughout the OrtSession lifetime?
   //  What sort of access is needed to model_path : read or read/write?
   OrtStatus*(ORT_API_CALL* CreateSession)(_In_ const OrtEnv* env, _In_ const ORTCHAR_T* model_path,
@@ -432,22 +432,22 @@ struct OrtApi {
   OrtStatus*(ORT_API_CALL* GetTensorMutableData)(_Inout_ OrtValue* value, _Outptr_ void** out)NO_EXCEPTION;
 
   /**
-	 * \param value A tensor created from OrtCreateTensor... function.
+	 * \param value A tensor created from CreateTensor... function.
 	 * \param s each A string array. Each string in this array must be null terminated.
 	 * \param s_len length of s
 	 */
   OrtStatus*(ORT_API_CALL* FillStringTensor)(_Inout_ OrtValue* value, _In_ const char* const* s, size_t s_len)NO_EXCEPTION;
 
   /**
-	 * \param value A tensor created from OrtCreateTensor... function.
+	 * \param value A tensor created from CreateTensor... function.
 	 * \param len total data length, not including the trailing '\0' chars.
 	 */
   OrtStatus*(ORT_API_CALL* GetStringTensorDataLength)(_In_ const OrtValue* value, _Out_ size_t* len)NO_EXCEPTION;
 
   /**
 	 * \param s string contents. Each string is NOT null-terminated.
-	 * \param value A tensor created from OrtCreateTensor... function.
-	 * \param s_len total data length, get it from OrtGetStringTensorDataLength
+	 * \param value A tensor created from CreateTensor... function.
+	 * \param s_len total data length, get it from GetStringTensorDataLength
 	 */
   OrtStatus*(ORT_API_CALL* GetStringTensorContent)(_In_ const OrtValue* value, _Out_ void* s, size_t s_len,
                                                    _Out_ size_t* offsets, size_t offsets_len)NO_EXCEPTION;
@@ -470,7 +470,7 @@ struct OrtApi {
   OrtStatus*(ORT_API_CALL* SetTensorElementType)(_Inout_ OrtTensorTypeAndShapeInfo*, enum ONNXTensorElementDataType type)NO_EXCEPTION;
 
   /**
- * \param info Created from OrtCreateTensorTypeAndShapeInfo() function
+ * \param info Created from CreateTensorTypeAndShapeInfo() function
  * \param dim_values An array with length of `dim_count`. Its elements can contain negative values.
  * \param dim_count length of dim_values
  */
diff --git a/onnxruntime/core/framework/data_transfer.cc b/onnxruntime/core/framework/data_transfer.cc
index 465655f03d6ec..80099c2e4e6fe 100644
--- a/onnxruntime/core/framework/data_transfer.cc
+++ b/onnxruntime/core/framework/data_transfer.cc
@@ -9,6 +9,14 @@ common::Status IDataTransfer::CopyTensor(const Tensor& src, Tensor& dst) const {
   return CopyTensor(src, dst, 0);
 }
 
+common::Status IDataTransfer::CopyTensors(const Tensor* src, Tensor* dst, int size) const {
+  ORT_ENFORCE(nullptr != src && nullptr != dst);
+  for (int i = 0; i < size; ++i) {
+    ORT_RETURN_IF_ERROR(CopyTensor(src[i], dst[i], 0));
+  }
+  return Status::OK();
+}
+
 bool CPUDataTransfer::CanCopy(const OrtDevice& src_device, const OrtDevice& dst_device) const {
   return src_device.Type() == OrtDevice::CPU && dst_device.Type() == OrtDevice::CPU;
 }
diff --git a/onnxruntime/core/framework/data_transfer.h b/onnxruntime/core/framework/data_transfer.h
index 1e707b196babc..798525149bff7 100644
--- a/onnxruntime/core/framework/data_transfer.h
+++ b/onnxruntime/core/framework/data_transfer.h
@@ -17,6 +17,7 @@ class IDataTransfer {
 
   virtual common::Status CopyTensor(const Tensor& src, Tensor& dst) const;
   virtual common::Status CopyTensor(const Tensor& src, Tensor& dst, int exec_queue_id) const = 0;
+  virtual common::Status CopyTensors(const Tensor* src, Tensor* dst, int size) const;
 };
 
 class CPUDataTransfer : public IDataTransfer {
diff --git a/onnxruntime/core/session/onnxruntime_c_api.cc b/onnxruntime/core/session/onnxruntime_c_api.cc
index 3a3dab335bcfa..99ff0212ad8ab 100644
--- a/onnxruntime/core/session/onnxruntime_c_api.cc
+++ b/onnxruntime/core/session/onnxruntime_c_api.cc
@@ -807,16 +807,8 @@ ORT_API_STATUS_IMPL(OrtApis::AllocatorGetInfo, _In_ const OrtAllocator* ptr, _Ou
   API_IMPL_END
 }
 
-///////////////////////////////////////////////////////////////////////////
-// Code to handle non-tensor types
-// OrtGetValueCount
-// OrtGetVaue
-// OrtCreateValue
-///////////////////////////////////////////////////////////////////////////
 const int NUM_MAP_INDICES = 2;
 
-////////////////////
-// OrtGetValueCount
 template <typename T>
 OrtStatus* OrtGetNumSequenceElements(const OrtValue* p_ml_value, size_t* out) {
   auto& data = p_ml_value->Get<T>();
@@ -867,7 +859,7 @@ ORT_API_STATUS_IMPL(OrtApis::GetValueCount, const OrtValue* value, size_t* out)
 }
 
 ///////////////////
-// OrtGetValue
+// OrtGetValueImplSeqOfMap
 template <typename T>
 static OrtStatus* OrtGetValueImplSeqOfMap(const OrtValue* p_ml_value, int index, OrtValue** out) {
   using TKey = typename T::value_type::key_type;

From 31200ed92ca5c6e26e4615a4c161ef04a3868041 Mon Sep 17 00:00:00 2001
From: George Wu <George.Wu@microsoft.com>
Date: Fri, 10 Jan 2020 08:40:40 -0800
Subject: [PATCH 40/75] speed up Windows TRT CI (#2811)

* don't run cuda tests if building with tensorrt

* remove unnecessary build options for win trt ci

* refactor win gpu tensorrt ci yml

* --numpy_version=1.17

* update

* update

* azcopy and cuda path
---
 tools/ci_build/build.py                       |   2 +-
 .../win-gpu-tensorrt-ci-pipeline.yml          | 120 +++++++++++++++---
 2 files changed, 106 insertions(+), 16 deletions(-)

diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index 09995aa63de8a..c7aad8e0e2760 100755
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -939,7 +939,7 @@ def main():
               else:
                 tensorrt_run_onnx_tests(build_dir, configs, "")
 
-            if args.use_cuda:
+            if args.use_cuda and not args.use_tensorrt:
               run_onnx_tests(build_dir, configs, onnx_test_data_dir, 'cuda', args.enable_multi_device_test, False, 2)           
 
             if args.use_ngraph:
diff --git a/tools/ci_build/github/azure-pipelines/win-gpu-tensorrt-ci-pipeline.yml b/tools/ci_build/github/azure-pipelines/win-gpu-tensorrt-ci-pipeline.yml
index d5401cd13b2e4..8c6db895c45ba 100644
--- a/tools/ci_build/github/azure-pipelines/win-gpu-tensorrt-ci-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/win-gpu-tensorrt-ci-pipeline.yml
@@ -1,16 +1,106 @@
 jobs:
-- template: templates/win-ci.yml
-  parameters:
-    AgentPool : 'Win-GPU-CUDA10'
-    DoDebugBuild: 'true'
-    DoCompliance: 'false'
-    BuildCommand: '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --enable_pybind --use_dnnl --use_tensorrt --tensorrt_home="C:\local\TensorRT-6.0.1.5" --build_shared_lib --build_csharp --enable_onnx_tests --use_cuda --cuda_version=10.0 --cuda_home="C:\local\cuda_10.0.130_win10" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda" --gen_doc'
-    JobName: 'Windows_CI_GPU_Dev'
-    DoNugetPack:  'false'
-    NuPackScript : ''
-    DoTestCoverage: 'false'
-    BuildArch: 'amd64'
-    SetVcvars: 'true'
-    MsbuildArguments: '/m /p:CudaToolkitDir=C:\local\cuda_10.0.130_win10\'
-    EnvSetupScript: 'setup_env_cuda.bat'
-    CudaVersion: '10.0'
+- job: 'build'
+  pool: 'Win-GPU-CUDA10'
+  strategy:
+    maxParallel: 2
+    matrix:
+      debug:
+        BuildConfig: 'Debug'
+      release:
+        BuildConfig: 'RelWithDebInfo'
+  variables:
+    OrtPackageId: 'Microsoft.ML.OnnxRuntime'
+    MsbuildArguments: '-detailedsummary -maxcpucount -consoleloggerparameters:PerformanceSummary'
+    OnnxRuntimeBuildDirectory: '$(Build.BinariesDirectory)'
+    DOTNET_SKIP_FIRST_TIME_EXPERIENCE: true
+    EnvSetupScript: setup_env.bat
+    buildArch: x64
+    setVcvars: true
+  timeoutInMinutes: 120
+  workspace:
+    clean: all
+  steps:
+  - task: UsePythonVersion@0
+    inputs:
+      versionSpec: '3.7'
+      addToPath: true
+      architecture: $(buildArch)
+
+  - task: BatchScript@1
+    displayName: 'setup env'
+    inputs:
+      filename: '$(Build.SourcesDirectory)\tools\ci_build\github\windows\$(EnvSetupScript)'
+      modifyEnvironment: true
+      workingFolder: '$(Build.BinariesDirectory)'
+
+  - task: PowerShell@1
+    displayName: 'Set CUDA path'
+    inputs:
+      scriptName: 'tools/ci_build/github/windows/set_cuda_path.ps1'
+      arguments: '-CudaMsbuildPath C:\local\cudaMsbuildIntegration-10.0.130-win10 -CudaVersion 10.0'
+
+  - task: PowerShell@2
+    displayName: 'Download AzCopy (used for download test data script)'
+    inputs:
+      targetType: 'inline'
+      script: |
+        Invoke-WebRequest -OutFile $(Build.BinariesDirectory)\azcopy.exe https://onnxruntimetestdata.blob.core.windows.net/models/azcopy.exe
+
+  - script: |
+     python -m pip install -q pyopenssl setuptools wheel numpy
+    workingDirectory: '$(Build.BinariesDirectory)'
+    displayName: 'Install python modules'
+
+  - task: PythonScript@0
+    displayName: 'Generate cmake config'
+    inputs:
+      scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
+      arguments: '--config $(BuildConfig) --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --update --numpy_version=1.17.4 --build_wheel --use_tensorrt --tensorrt_home="C:\local\TensorRT-6.0.1.5" --enable_onnx_tests --cuda_version=10.0 --cuda_home="C:\local\cuda_10.0.130_win10" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"'
+      workingDirectory: '$(Build.BinariesDirectory)'
+
+  - task: VSBuild@1
+    displayName: 'Build'
+    inputs:
+      solution: '$(Build.BinariesDirectory)\$(BuildConfig)\onnxruntime.sln'
+      platform: 'x64'
+      configuration: $(BuildConfig)
+      msbuildArgs: $(MsbuildArguments)
+      msbuildArchitecture: $(buildArch)
+      maximumCpuCount: true
+      logProjectEvents: false
+      workingFolder: '$(Build.BinariesDirectory)\$(BuildConfig)'
+      createLogFile: true
+
+  - task: PythonScript@0
+    displayName: 'Build wheel'
+    inputs:
+      scriptPath: '$(Build.SourcesDirectory)\setup.py'
+      arguments: 'bdist_wheel'
+      workingDirectory: '$(Build.BinariesDirectory)\$(BuildConfig)\$(BuildConfig)'
+
+  - template: templates/set-test-data-variables-step.yml
+
+  - task: PythonScript@0
+    displayName: 'Download test data'
+    inputs:
+      scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\github\download_test_data.py'
+      arguments: --test_data_url $(TestDataUrl) --build_dir $(Build.BinariesDirectory)
+      workingDirectory: $(Build.BinariesDirectory)
+
+  - script: |
+     mklink  /D /J $(Build.BinariesDirectory)\$(BuildConfig)\models $(Build.BinariesDirectory)\models
+     DIR dist\ /S /B > wheel_filename_file
+     set /p WHEEL_FILENAME=<wheel_filename_file
+     del wheel_filename_file
+     python.exe -m pip install -q --upgrade %WHEEL_FILENAME%
+     set PATH=$(Build.BinariesDirectory)\$(BuildConfig)\$(BuildConfig);%PATH%
+     python $(Build.SourcesDirectory)\tools\ci_build\build.py --config $(BuildConfig) --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --test --numpy_version=1.17.4 --build_wheel --enable_onnx_tests --use_tensorrt --cuda_version=10.0 --tensorrt_home="C:\local\TensorRT-6.0.1.5" --cuda_home="C:\local\cuda_10.0.130_win10" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"
+
+    workingDirectory: '$(Build.BinariesDirectory)\$(BuildConfig)\$(BuildConfig)'
+    displayName: 'Run tests'
+
+
+  - task: ms.vss-governance-buildtask.governance-build-task-component-detection.ComponentGovernanceComponentDetection@0
+    displayName: 'Component Detection'
+    condition: succeeded()
+

From 48e042868f67a241e878e3cbaad938de39902df4 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 10 Jan 2020 10:52:23 -0800
Subject: [PATCH 41/75] Update test data (#2356)

---
 .../InferenceTest.cs                          | 133 ++++++------
 onnxruntime/test/onnx/main.cc                 | 196 +++++++++---------
 .../set-test-data-variables-step.yml          |  13 +-
 3 files changed, 168 insertions(+), 174 deletions(-)

diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
index 29d3181568cd3..e60cbb8aacf7f 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
@@ -328,7 +328,8 @@ private static Dictionary<string, string> GetSkippedModels()
                 { "mlperf_ssd_mobilenet_300", "Could not find file output_0.pb" },
                 { "tf_resnet_v1_50", "result mismatch when Conv BN Fusion is applied" },
                 { "tf_resnet_v1_101", "result mismatch when Conv BN Fusion is applied" },
-                { "tf_resnet_v1_152", "result mismatch when Conv BN Fusion is applied" }
+                { "tf_resnet_v1_152", "result mismatch when Conv BN Fusion is applied" },
+                { "mask_rcnn_keras", "Model should be edited to remove the extra outputs" },
             };
 
             // The following models fails on nocontribops win CI
@@ -437,84 +438,86 @@ private void TestPreTrainedModels(string opset, string modelName)
                     {
                         testDataDirNamePattern = "seq_lens*"; // discrepency in data directory
                     }
-                    var testDataDir = modelDir.EnumerateDirectories(testDataDirNamePattern).First();
-                    var inputContainer = new List<NamedOnnxValue>();
-                    var outputContainer = new List<NamedOnnxValue>();
-                    foreach (var f in testDataDir.EnumerateFiles("input_*.pb"))
-                    {
-                        inputContainer.Add(LoadTensorFromFilePb(f.FullName, inMeta));
-                    }
-                    foreach (var f in testDataDir.EnumerateFiles("output_*.pb"))
+                    foreach (var testDataDir in modelDir.EnumerateDirectories(testDataDirNamePattern))
                     {
-                        outputContainer.Add(LoadTensorFromFilePb(f.FullName, session.OutputMetadata));
-                    }
+                        var inputContainer = new List<NamedOnnxValue>();
+                        var outputContainer = new List<NamedOnnxValue>();
+                        foreach (var f in testDataDir.EnumerateFiles("input_*.pb"))
+                        {
+                            inputContainer.Add(LoadTensorFromFilePb(f.FullName, inMeta));
+                        }
+                        foreach (var f in testDataDir.EnumerateFiles("output_*.pb"))
+                        {
+                            outputContainer.Add(LoadTensorFromFilePb(f.FullName, session.OutputMetadata));
+                        }
 
-                    using (var resultCollection = session.Run(inputContainer))
-                    {
-                        foreach (var result in resultCollection)
+                        using (var resultCollection = session.Run(inputContainer))
                         {
-                            Assert.True(session.OutputMetadata.ContainsKey(result.Name));
-                            var outputMeta = session.OutputMetadata[result.Name];
-                            NamedOnnxValue outputValue = null;
-                            foreach (var o in outputContainer)
-                            {
-                                if (o.Name == result.Name)
-                                {
-                                    outputValue = o;
-                                    break;
-                                }
-                            }
-                            if (outputValue == null)
-                            {
-                                outputValue = outputContainer.First(); // in case the output data file does not contain the name
-                            }
-                            if (outputMeta.IsTensor)
+                            foreach (var result in resultCollection)
                             {
-                                if (outputMeta.ElementType == typeof(float))
-                                {
-                                    Assert.Equal(result.AsTensor<float>(), outputValue.AsTensor<float>(), new floatComparer());
-                                }
-                                else if (outputMeta.ElementType == typeof(int))
-                                {
-                                    Assert.Equal(result.AsTensor<int>(), outputValue.AsTensor<int>(), new ExactComparer<int>());
-                                }
-                                else if (outputMeta.ElementType == typeof(uint))
-                                {
-                                    Assert.Equal(result.AsTensor<uint>(), outputValue.AsTensor<uint>(), new ExactComparer<uint>());
-                                }
-                                else if (outputMeta.ElementType == typeof(short))
+                                Assert.True(session.OutputMetadata.ContainsKey(result.Name));
+                                var outputMeta = session.OutputMetadata[result.Name];
+                                NamedOnnxValue outputValue = null;
+                                foreach (var o in outputContainer)
                                 {
-                                    Assert.Equal(result.AsTensor<short>(), outputValue.AsTensor<short>(), new ExactComparer<short>());
+                                    if (o.Name == result.Name)
+                                    {
+                                        outputValue = o;
+                                        break;
+                                    }
                                 }
-                                else if (outputMeta.ElementType == typeof(ushort))
+                                if (outputValue == null)
                                 {
-                                    Assert.Equal(result.AsTensor<ushort>(), outputValue.AsTensor<ushort>(), new ExactComparer<ushort>());
+                                    outputValue = outputContainer.First(); // in case the output data file does not contain the name
                                 }
-                                else if (outputMeta.ElementType == typeof(long))
+                                if (outputMeta.IsTensor)
                                 {
-                                    Assert.Equal(result.AsTensor<long>(), outputValue.AsTensor<long>(), new ExactComparer<long>());
-                                }
-                                else if (outputMeta.ElementType == typeof(ulong))
-                                {
-                                    Assert.Equal(result.AsTensor<ulong>(), outputValue.AsTensor<ulong>(), new ExactComparer<ulong>());
-                                }
-                                else if (outputMeta.ElementType == typeof(byte))
-                                {
-                                    Assert.Equal(result.AsTensor<byte>(), outputValue.AsTensor<byte>(), new ExactComparer<byte>());
-                                }
-                                else if (outputMeta.ElementType == typeof(bool))
-                                {
-                                    Assert.Equal(result.AsTensor<bool>(), outputValue.AsTensor<bool>(), new ExactComparer<bool>());
+                                    if (outputMeta.ElementType == typeof(float))
+                                    {
+                                        Assert.Equal(result.AsTensor<float>(), outputValue.AsTensor<float>(), new floatComparer());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(int))
+                                    {
+                                        Assert.Equal(result.AsTensor<int>(), outputValue.AsTensor<int>(), new ExactComparer<int>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(uint))
+                                    {
+                                        Assert.Equal(result.AsTensor<uint>(), outputValue.AsTensor<uint>(), new ExactComparer<uint>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(short))
+                                    {
+                                        Assert.Equal(result.AsTensor<short>(), outputValue.AsTensor<short>(), new ExactComparer<short>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(ushort))
+                                    {
+                                        Assert.Equal(result.AsTensor<ushort>(), outputValue.AsTensor<ushort>(), new ExactComparer<ushort>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(long))
+                                    {
+                                        Assert.Equal(result.AsTensor<long>(), outputValue.AsTensor<long>(), new ExactComparer<long>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(ulong))
+                                    {
+                                        Assert.Equal(result.AsTensor<ulong>(), outputValue.AsTensor<ulong>(), new ExactComparer<ulong>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(byte))
+                                    {
+                                        Assert.Equal(result.AsTensor<byte>(), outputValue.AsTensor<byte>(), new ExactComparer<byte>());
+                                    }
+                                    else if (outputMeta.ElementType == typeof(bool))
+                                    {
+                                        Assert.Equal(result.AsTensor<bool>(), outputValue.AsTensor<bool>(), new ExactComparer<bool>());
+                                    }
+                                    else
+                                    {
+                                        Assert.True(false, "The TestPretrainedModels does not yet support output of type " + nameof(outputMeta.ElementType));
+                                    }
                                 }
                                 else
                                 {
-                                    Assert.True(false, "The TestPretrainedModels does not yet support output of type " + nameof(outputMeta.ElementType));
+                                    Assert.True(false, "TestPretrainedModel cannot handle non-tensor outputs yet");
                                 }
                             }
-                            else
-                            {
-                                Assert.True(false, "TestPretrainedModel cannot handle non-tensor outputs yet");
-                            }
                         }
                     }
                 }
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index 6fa0b30456d03..ff881510714b4 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -357,7 +357,7 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
     static const ORTCHAR_T* cuda_flaky_tests[] = {
         ORT_TSTR("fp16_inception_v1"),
         ORT_TSTR("fp16_shufflenet"), ORT_TSTR("fp16_tiny_yolov2")};
-    static const ORTCHAR_T* dml_disabled_tests[] = {ORT_TSTR("mlperf_ssd_resnet34_1200"), ORT_TSTR("mlperf_ssd_mobilenet_300"), ORT_TSTR("mask_rcnn_keras"), ORT_TSTR("mask_rcnn"), ORT_TSTR("faster_rcnn")};
+    static const ORTCHAR_T* dml_disabled_tests[] = {ORT_TSTR("mlperf_ssd_resnet34_1200"), ORT_TSTR("mlperf_ssd_mobilenet_300"), ORT_TSTR("mask_rcnn"), ORT_TSTR("faster_rcnn")};
     static const ORTCHAR_T* dnnl_disabled_tests[] = {ORT_TSTR("test_densenet121"), ORT_TSTR("test_resnet18v2"), ORT_TSTR("test_resnet34v2"), ORT_TSTR("test_resnet50v2"), ORT_TSTR("test_resnet101v2"),
                                                      ORT_TSTR("test_resnet101v2"), ORT_TSTR("test_vgg19"), ORT_TSTR("tf_inception_resnet_v2"), ORT_TSTR("tf_inception_v1"), ORT_TSTR("tf_inception_v3"), ORT_TSTR("tf_inception_v4"), ORT_TSTR("tf_mobilenet_v1_1.0_224"),
                                                      ORT_TSTR("tf_mobilenet_v2_1.0_224"), ORT_TSTR("tf_mobilenet_v2_1.4_224"), ORT_TSTR("tf_nasnet_large"), ORT_TSTR("tf_pnasnet_large"), ORT_TSTR("tf_resnet_v1_50"), ORT_TSTR("tf_resnet_v1_101"), ORT_TSTR("tf_resnet_v1_101"),
@@ -398,25 +398,22 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
     std::string res = stat.ToString();
     fwrite(res.c_str(), 1, res.size(), stdout);
   }
-  // clang-format off
 
-  struct BrokenTest
-  {
+  struct BrokenTest {
     std::string test_name_;
     std::string reason_;
-    std::set<std::string> broken_versions_ = {}; // apply to all versions if empty
+    std::set<std::string> broken_versions_ = {};  // apply to all versions if empty
     BrokenTest(std::string name, std::string reason) : test_name_(std::move(name)), reason_(std::move(reason)) {}
-    BrokenTest(std::string name, std::string reason, const std::initializer_list<std::string>& versions) :
-      test_name_(std::move(name)), reason_(std::move(reason)), broken_versions_(versions) {}
-    bool operator < (const struct BrokenTest& test) const {
-        return strcmp(test_name_.c_str(), test.test_name_.c_str()) < 0;
+    BrokenTest(std::string name, std::string reason, const std::initializer_list<std::string>& versions) : test_name_(std::move(name)), reason_(std::move(reason)), broken_versions_(versions) {}
+    bool operator<(const struct BrokenTest& test) const {
+      return strcmp(test_name_.c_str(), test.test_name_.c_str()) < 0;
     }
   };
 
   std::set<BrokenTest> broken_tests = {
       {"BERT_Squad", "test data bug"},
-      {"constantofshape_float_ones", "test data bug", {"onnx141","onnx150"}},
-      {"constantofshape_int_zeros", "test data bug", {"onnx141","onnx150"}},
+      {"constantofshape_float_ones", "test data bug", {"onnx141", "onnx150"}},
+      {"constantofshape_int_zeros", "test data bug", {"onnx141", "onnx150"}},
       {"convtranspose_3d", "3d convtranspose not supported yet"},
       {"cast_STRING_to_FLOAT", "Linux CI has old ONNX python package with bad test data", {"onnx141"}},
       // Numpy float to string has unexpected rounding for some results given numpy default precision is meant to be 8.
@@ -443,102 +440,105 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
       {"resize_upsample_sizes_nearest_round_prefer_ceil_asymmetric", "Bad onnx test output. Needs test fix."},
       {"bitshift_right_uint16", "BitShift(11) uint16 support not enabled currently"},
       {"bitshift_left_uint16", "BitShift(11) uint16 support not enabled currently"},
-      {"maxunpool_export_with_output_shape", "Invalid output in ONNX test. See https://github.com/onnx/onnx/issues/2398" },
-};
-
-#ifdef USE_NGRAPH
-  broken_tests.insert({"qlinearconv", "ambiguity in scalar dimensions [] vs [1]"});
-  broken_tests.insert({"clip_splitbounds", "not implemented yet for opset 11"});
-  broken_tests.insert({"clip_outbounds", "not implemented yet for opset 11"});
-  broken_tests.insert({"clip_example", "not implemented yet for opset 11"});
-  broken_tests.insert({"clip_default_min", "not implemented yet for opset 11"});
-  broken_tests.insert({"clip_default_max", "not implemented yet for opset 11"});
-  broken_tests.insert({"clip", "not implemented yet for opset 11"});
-  broken_tests.insert({"depthtospace_crd_mode_example", "NGraph does not support CRD mode"});
-  broken_tests.insert({"depthtospace_crd_mode", "NGraph does not support CRD mode"});
-  broken_tests.insert({"gemm_default_no_bias", "not implemented yet for opset 11"});
-  broken_tests.insert({"quantizelinear", "ambiguity in scalar dimensions [] vs [1]", {"onnx150"}});
-  broken_tests.insert({"dequantizelinear", "ambiguity in scalar dimensions [] vs [1]", {"onnx150"}});
-  broken_tests.insert({"mlperf_ssd_resnet34_1200", "Results mismatch"});
-  broken_tests.insert({"BERT_Squad", "Invalid Feed Input Name:input4"});
-  broken_tests.insert({"mask_rcnn_keras", "Results mismatch: 8 of 81000"});
-  broken_tests.insert({"candy", "Results mismatch: 2 of 150528"});
-  broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
-#endif
+      {"maxunpool_export_with_output_shape", "Invalid output in ONNX test. See https://github.com/onnx/onnx/issues/2398"},
+  };
 
-#ifdef USE_DNNL
-  broken_tests.insert({"tf_mobilenet_v2_1.0_224", "result mismatch"});
-  broken_tests.insert({"tf_mobilenet_v2_1.4_224", "result mismatch"});
-  broken_tests.insert({"tf_mobilenet_v1_1.0_224", "result mismatch"});
-  broken_tests.insert({"mobilenetv2-1.0", "result mismatch"});
-  broken_tests.insert({"candy", "result mismatch"});
-  broken_tests.insert({"range_float_type_positive_delta_expanded", "get unknown exception from DNNL EP"});
-  broken_tests.insert({"range_int32_type_negative_delta_expanded", "get unknown exception from DNNL EP"});
-  broken_tests.insert({"averagepool_2d_ceil", "maxpool ceiling not supported"});
-  broken_tests.insert({"maxpool_2d_ceil", "maxpool ceiling not supported"});
-  broken_tests.insert({"maxpool_2d_dilations", "maxpool dilations not supported"});
-  broken_tests.insert({"mlperf_ssd_resnet34_1200", "test pass on dev box but fails on CI build"}); 
-  broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
-#endif
+  if (enable_ngraph) {
+    broken_tests.insert({"qlinearconv", "ambiguity in scalar dimensions [] vs [1]"});
+    broken_tests.insert({"clip_splitbounds", "not implemented yet for opset 11"});
+    broken_tests.insert({"clip_outbounds", "not implemented yet for opset 11"});
+    broken_tests.insert({"clip_example", "not implemented yet for opset 11"});
+    broken_tests.insert({"clip_default_min", "not implemented yet for opset 11"});
+    broken_tests.insert({"clip_default_max", "not implemented yet for opset 11"});
+    broken_tests.insert({"clip", "not implemented yet for opset 11"});
+    broken_tests.insert({"depthtospace_crd_mode_example", "NGraph does not support CRD mode"});
+    broken_tests.insert({"depthtospace_crd_mode", "NGraph does not support CRD mode"});
+    broken_tests.insert({"gemm_default_no_bias", "not implemented yet for opset 11"});
+    broken_tests.insert({"quantizelinear", "ambiguity in scalar dimensions [] vs [1]", {"onnx150"}});
+    broken_tests.insert({"dequantizelinear", "ambiguity in scalar dimensions [] vs [1]", {"onnx150"}});
+    broken_tests.insert({"mlperf_ssd_resnet34_1200", "Results mismatch"});
+    broken_tests.insert({"BERT_Squad", "Invalid Feed Input Name:input4"});
+    broken_tests.insert({"candy", "Results mismatch: 2 of 150528"});
+    broken_tests.insert({"tf_mobilenet_v1_1.0_224", "Results mismatch"});    
+    broken_tests.insert({"tf_mobilenet_v2_1.0_224", "Results mismatch"});
+    broken_tests.insert({"tf_mobilenet_v2_1.4_224", "Results mismatch"});
+    broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
+  }
+  if (enable_nuphar) {
+    broken_tests.insert({"cgan", "TVM exception during initialization"});
+  }
+  if (enable_dnnl) {
+    broken_tests.insert({"tf_mobilenet_v2_1.0_224", "result mismatch"});
+    broken_tests.insert({"tf_mobilenet_v2_1.4_224", "result mismatch"});
+    broken_tests.insert({"tf_mobilenet_v1_1.0_224", "result mismatch"});
+    broken_tests.insert({"mobilenetv2-1.0", "result mismatch"});
+    broken_tests.insert({"candy", "result mismatch"});
+    broken_tests.insert({"range_float_type_positive_delta_expanded", "get unknown exception from DNNL EP"});
+    broken_tests.insert({"range_int32_type_negative_delta_expanded", "get unknown exception from DNNL EP"});
+    broken_tests.insert({"averagepool_2d_ceil", "maxpool ceiling not supported"});
+    broken_tests.insert({"maxpool_2d_ceil", "maxpool ceiling not supported"});
+    broken_tests.insert({"maxpool_2d_dilations", "maxpool dilations not supported"});
+    broken_tests.insert({"mlperf_ssd_resnet34_1200", "test pass on dev box but fails on CI build"});
+    broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
+  }
 
-#ifdef USE_OPENVINO
-  broken_tests.insert({"fp16_shufflenet", "accuracy mismatch with fp16 precision"});
-  broken_tests.insert({"fp16_inception_v1", "accuracy mismatch with fp16 precision"});
-  broken_tests.insert({"fp16_tiny_yolov2", "accuaracy mismatch with fp16 precision"});
-  broken_tests.insert({"scan_sum", "disable temporarily"});
-  broken_tests.insert({"scan9_sum", "disable temporarily"});
-  broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
+  if (enable_openvino) {
+    broken_tests.insert({"fp16_shufflenet", "accuracy mismatch with fp16 precision"});
+    broken_tests.insert({"fp16_inception_v1", "accuracy mismatch with fp16 precision"});
+    broken_tests.insert({"fp16_tiny_yolov2", "accuaracy mismatch with fp16 precision"});
+    broken_tests.insert({"scan_sum", "disable temporarily"});
+    broken_tests.insert({"scan9_sum", "disable temporarily"});
+    broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
 #ifdef OPENVINO_CONFIG_GPU_FP32
-  broken_tests.insert({"tiny_yolov2", "accuracy mismatch"});
-  broken_tests.insert({"div", "will be fixed in the next release"});
+    broken_tests.insert({"tiny_yolov2", "accuracy mismatch"});
+    broken_tests.insert({"div", "will be fixed in the next release"});
 #ifdef OPENVINO_CONFIG_GPU_FP16
-  broken_tests.insert({"div", "will be fixed in the next release"});
-#endif
+    broken_tests.insert({"div", "will be fixed in the next release"});
 #endif
 #endif
+  }
 
-#ifdef USE_NNAPI
-  broken_tests.insert({"scan9_sum", "Error with the extra graph"});
-  broken_tests.insert({"scan_sum", "Error with the extra graph"});
-  broken_tests.insert({"mvn_expanded", "Failed to find kernel for MemcpyFromHost(1) (node Memcpy_1)"});
-  broken_tests.insert({"dynamicquantizelinear_expanded", "Temporarily disabled pending investigation"});
-  broken_tests.insert({"dynamicquantizelinear_max_adjusted_expanded", "Temporarily disabled pending investigation"});
-  broken_tests.insert({"dynamicquantizelinear_min_adjusted_expanded", "Temporarily disabled pending investigation"});
-  broken_tests.insert({"gemm_transposeB", "Temporarily disabled pending investigation"});
-  broken_tests.insert({"range_float_type_positive_delta_expanded", "Temporarily disabled pending investigation"});
-  broken_tests.insert({"range_int32_type_negative_delta_expanded", "Temporarily disabled pending investigation"});
-  broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
-#endif
+  if (enable_nnapi) {
+    broken_tests.insert({"scan9_sum", "Error with the extra graph"});
+    broken_tests.insert({"scan_sum", "Error with the extra graph"});
+    broken_tests.insert({"mvn_expanded", "Failed to find kernel for MemcpyFromHost(1) (node Memcpy_1)"});
+    broken_tests.insert({"dynamicquantizelinear_expanded", "Temporarily disabled pending investigation"});
+    broken_tests.insert({"dynamicquantizelinear_max_adjusted_expanded", "Temporarily disabled pending investigation"});
+    broken_tests.insert({"dynamicquantizelinear_min_adjusted_expanded", "Temporarily disabled pending investigation"});
+    broken_tests.insert({"gemm_transposeB", "Temporarily disabled pending investigation"});
+    broken_tests.insert({"range_float_type_positive_delta_expanded", "Temporarily disabled pending investigation"});
+    broken_tests.insert({"range_int32_type_negative_delta_expanded", "Temporarily disabled pending investigation"});
+    broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
+  }
 
-#ifdef USE_TENSORRT
-  broken_tests.insert({"fp16_shufflenet", "TRT EP bug"});
-  broken_tests.insert({"fp16_inception_v1", "TRT EP bug"});
-  broken_tests.insert({"fp16_tiny_yolov2", "TRT EP bug"});
-  broken_tests.insert({"tf_inception_v3", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_mobilenet_v1_1.0_224", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_mobilenet_v2_1.0_224", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_mobilenet_v2_1.4_224", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_resnet_v1_101", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_resnet_v1_152", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_resnet_v1_50", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_resnet_v2_101", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_resnet_v2_152", "TRT Engine couldn't be created"});
-  broken_tests.insert({"tf_resnet_v2_50", "TRT Engine couldn't be created"});
-  broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
-#endif
+  if (enable_tensorrt) {
+    broken_tests.insert({"fp16_shufflenet", "TRT EP bug"});
+    broken_tests.insert({"fp16_inception_v1", "TRT EP bug"});
+    broken_tests.insert({"fp16_tiny_yolov2", "TRT EP bug"});
+    broken_tests.insert({"tf_inception_v3", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_mobilenet_v1_1.0_224", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_mobilenet_v2_1.0_224", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_mobilenet_v2_1.4_224", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_resnet_v1_101", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_resnet_v1_152", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_resnet_v1_50", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_resnet_v2_101", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_resnet_v2_152", "TRT Engine couldn't be created"});
+    broken_tests.insert({"tf_resnet_v2_50", "TRT Engine couldn't be created"});
+    broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
+  }
 
-#ifdef USE_CUDA
-  broken_tests.insert({"candy", "result mismatch"});
-  broken_tests.insert({"mask_rcnn_keras", "result mismatch"});
-  broken_tests.insert({"mlperf_ssd_mobilenet_300", "unknown error"});
-  broken_tests.insert({"mlperf_ssd_resnet34_1200", "unknown error"});
-  broken_tests.insert({"tf_inception_v1", "flaky test"}); //TODO: Investigate cause for flakiness
-  broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
-#endif
+  if (enable_cuda) {
+    broken_tests.insert({"candy", "result mismatch"});
+    broken_tests.insert({"tinyyolov3", "The parameter is incorrect"});
+    broken_tests.insert({"mlperf_ssd_mobilenet_300", "unknown error"});
+    broken_tests.insert({"mlperf_ssd_resnet34_1200", "unknown error"});
+    broken_tests.insert({"tf_inception_v1", "flaky test"});  //TODO: Investigate cause for flakiness
+    broken_tests.insert({"convtranspose_1d", "1d convtranspose not supported yet"});
+  }
 
-#ifdef USE_DML
-  if (enable_dml)
-  {
+  if (enable_dml) {
+    broken_tests.insert({"tinyyolov3", "The parameter is incorrect"});
     broken_tests.insert({"PixelShuffle", "Test requires 6D Reshape, which isn't supported by DirectML"});
     broken_tests.insert({"operator_permute2", "Test requires 6D Transpose, which isn't supported by DirectML"});
     broken_tests.insert({"resize_downsample_linear", "ORT 0.4 uses asymmetric but will conform to half_pixel in the next ONNX version."});
@@ -562,8 +562,6 @@ int real_main(int argc, char* argv[], Ort::Env& env) {
     broken_tests.insert({"candy", "Temporarily disabled pending investigation"});
     broken_tests.insert({"BERT_Squad", "Temporarily disabled pending investigation"});
   }
-#endif
-  // clang-format on
 
 #if defined(_WIN32) && !defined(_WIN64)
   broken_tests.insert({"vgg19", "failed: bad allocation"});
diff --git a/tools/ci_build/github/azure-pipelines/templates/set-test-data-variables-step.yml b/tools/ci_build/github/azure-pipelines/templates/set-test-data-variables-step.yml
index 658e3515bd44c..abec88edb183f 100644
--- a/tools/ci_build/github/azure-pipelines/templates/set-test-data-variables-step.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/set-test-data-variables-step.yml
@@ -1,18 +1,11 @@
-# sets variables $(TestDataUrl) and $(TestDataChecksum)
+# sets variables $(TestDataUrl)
 
 parameters:
-  TestDataUrl: https://onnxruntimetestdata.blob.core.windows.net/models/20190925.zip
-  TestDataChecksum: ae5fb5e3dd5e4937c8343d13e3be680c
+  TestDataUrl: https://onnxruntimetestdata.blob.core.windows.net/models/20191107.zip
 
 steps:
 - task: CmdLine@1
   displayName: 'Set TestDataUrl variable'
   inputs:
     filename: echo
-    arguments: '##vso[task.setvariable variable=TestDataUrl;]${{parameters.TestDataUrl}}'
-
-- task: CmdLine@1
-  displayName: 'Set TestDataChecksum variable'
-  inputs:
-    filename: echo
-    arguments: '##vso[task.setvariable variable=TestDataChecksum;]${{parameters.TestDataChecksum}}'
+    arguments: '##vso[task.setvariable variable=TestDataUrl;]${{parameters.TestDataUrl}}'
\ No newline at end of file

From afa48b7e13ec2f706ec0569e24f57b30b9c604dd Mon Sep 17 00:00:00 2001
From: Dmitri Smirnov <yuslepukhin@users.noreply.github.com>
Date: Fri, 10 Jan 2020 13:27:51 -0800
Subject: [PATCH 42/75] Add timeseries imputer transformer featurizer kernel
 (#2813)

 Make kernels non-template. Add input constraint for learnt data.
  Fixup tests.
  Add two more featurizers along with tests. Tests fail.
  min_max_scalar_transformer
  robust_scalar_transformer
  Fix tests serialized stream by prepending version bytes.
  Add inputation_marker_transfomer and the test.
  Fix up float/double type designations.
 Added label_encoder_transformer along with a test.
  string_throw case is broken at the momement.
  Fix labelencodertransfomer_test.cc string_throw case
  Rename maxabsscalertransformer_test.cc
  Add MissingDummiesTransformer along with the test.
  Update manifest.
  Add TimeSeriesImputerTransformer definition, implementation and tests
---
 cgmanifest.json                               |   2 +-
 cmake/external/featurizers.cmake              |   3 +-
 .../graph/featurizers_ops/featurizers_defs.cc | 182 ++++++++++++-
 .../cpu/time_series_imputer_transformer.cc    | 244 ++++++++++++++++++
 .../cpu_featurizers_kernels.cc                |   2 +
 .../time_series_imputer_transformer_test.cc   | 196 ++++++++++++++
 6 files changed, 622 insertions(+), 7 deletions(-)
 create mode 100644 onnxruntime/featurizers_ops/cpu/time_series_imputer_transformer.cc
 create mode 100644 onnxruntime/test/featurizers_ops/time_series_imputer_transformer_test.cc

diff --git a/cgmanifest.json b/cgmanifest.json
index e2b2c73a60a57..62ad0a3031625 100644
--- a/cgmanifest.json
+++ b/cgmanifest.json
@@ -450,7 +450,7 @@
       {
          "component": {
             "git": {
-               "commitHash": "a11f5002af58a03d5902b13ef65c84cedb499024",
+               "commitHash": "573070aeeb77e267da2579ac1d75d92c688bbe97",
                "repositoryUrl": "https://github.com/microsoft/FeaturizersLibrary.git"
             },
             "type": "git"
diff --git a/cmake/external/featurizers.cmake b/cmake/external/featurizers.cmake
index acc9b25c564d1..4700e85f032db 100644
--- a/cmake/external/featurizers.cmake
+++ b/cmake/external/featurizers.cmake
@@ -3,7 +3,7 @@
 # This source code should not depend on the onnxruntime and may be built independently
 
 set(featurizers_URL "https://github.com/microsoft/FeaturizersLibrary.git")
-set(featurizers_TAG "a11f5002af58a03d5902b13ef65c84cedb499024")
+set(featurizers_TAG "573070aeeb77e267da2579ac1d75d92c688bbe97")
 
 set(featurizers_pref FeaturizersLibrary)
 set(featurizers_ROOT ${PROJECT_SOURCE_DIR}/external/${featurizers_pref})
@@ -24,6 +24,7 @@ if (WIN32)
             BINARY_DIR ${featurizers_BINARY_DIR}
             CMAKE_ARGS -Dfeaturizers_MSVC_STATIC_RUNTIME=${onnxruntime_MSVC_STATIC_RUNTIME}
             INSTALL_COMMAND ""
+            
         )
 else()
     ExternalProject_Add(featurizers_lib
diff --git a/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc b/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
index 5c12ef18bce8f..c46822774918d 100644
--- a/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
+++ b/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
@@ -41,6 +41,7 @@ static void RegisterMinMaxScalarFeaturizerVer1();
 static void RegisterMissingDummiesFeaturizerVer1();
 static void RegisterRobustScalarFeaturizerVer1();
 static void RegisterStringFeaturizerVer1();
+static void RegisterTimeSeriesImputerFeaturizerVer1();
 
 // ----------------------------------------------------------------------
 // ----------------------------------------------------------------------
@@ -55,6 +56,7 @@ void RegisterMSFeaturizersSchemas() {
   RegisterMissingDummiesFeaturizerVer1();
   RegisterRobustScalarFeaturizerVer1();
   RegisterStringFeaturizerVer1();
+  RegisterTimeSeriesImputerFeaturizerVer1();
 }
 
 // ----------------------------------------------------------------------
@@ -212,7 +214,7 @@ void RegisterDateTimeFeaturizerVer1() {
                 case 0:
                   propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, output);
                   break;
-                case 1: // fall through
+                case 1:  // fall through
                 case 2:
                 case 3:
                 case 4:
@@ -223,11 +225,11 @@ void RegisterDateTimeFeaturizerVer1() {
                 case 9:
                   propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
                   break;
-                case 10: // fall through
+                case 10:  // fall through
                 case 11:
                   propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT16, output);
                   break;
-                case 12: // fall through
+                case 12:  // fall through
                 case 13:
                 case 14:
                   propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
@@ -595,7 +597,6 @@ void RegisterRobustScalarFeaturizerVer1() {
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT32 ||
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_UINT64 ||
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_DOUBLE) {
-              ctx.getOutputType(0)->mutable_tensor_type()->set_elem_type(ONNX_NAMESPACE::TensorProto_DataType_DOUBLE);
               propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_DOUBLE, 0);
             } else {
               fail_type_inference("input 1 is expected to have a accepted type");
@@ -648,7 +649,178 @@ void RegisterStringFeaturizerVer1() {
       .TypeAndShapeInferenceFunction(
           [](ONNX_NAMESPACE::InferenceContext& ctx) {
             propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, 0);
-            propagateShapeFromInputToOutput(ctx, 1, 0);
+            if (hasInputShape(ctx, 1)) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
+          });
+}
+
+void RegisterTimeSeriesImputerFeaturizerVer1() {
+  static const char* doc = R"DOC(
+    Imputes rows and column values such that the generated output does not contain any
+    time gaps per grain (based on the time gaps encountered during training) and that
+    all missing column values are populated according to a strategy (forward fill,
+    backward fill, mode, etc.).
+
+    This Featurizer is unique in that it will produce 0:N rows per invocation, depending upon the
+    input data.
+
+    C++-style pseudo signature:
+      template <typename... GrainColValueTs, typename... DataColValueTs>
+      std::vector<
+        std::tuple<
+          bool, // true if the row was added
+          std::chrono::system_clock::time_point,
+          std::tuple<GrainColValueTs...>,
+          std::tuple<DataColValueTs...>
+        >
+      > execute(
+        std::chrono::system_clock::time_point const &value,
+        std::tuple<GrainColValueTs...> const &grain,
+        std::tuple<DataColValueTs...> const &colData
+      );
+
+    Examples:
+      During training, the time period was found to be 1 day...
+
+      Input:
+        +------+-------+------------------+-------------------+
+        | time | grain | forward fill col | backward fill col |
+        +======+=======+==================+===================+
+        | 1    | A     | 10               | None              |
+        +------+-------+------------------+-------------------+
+        | 2    | A     | None             | 200               |
+        +------+-------+------------------+-------------------+
+        | 1    | B     | -10              | -100              |
+        +------+-------+------------------+-------------------+
+        | 4    | A     | 40               | 400               |
+        +------+-------+------------------+-------------------+
+        | 6    | A     | 60               | 600               |
+        +------+-------+------------------+-------------------+
+        | 3    | B     | -30              | -300              |
+        +------+-------+------------------+-------------------+
+
+      Output:
+        +-------+------+-------+------------------+-------------------+
+        | Added | time | grain | forward fill col | backward fill col |
+        +=======+======+=======+==================+===================+
+        | false | 1    | A     | 10               | 200 (from 2)      |
+        +-------+------+-------+------------------+-------------------+
+        | false | 2    | A     | 10 (from 1)      | 200               |
+        +-------+------+-------+------------------+-------------------+
+        | true  | 3    | A     | 10 (from 2)      | 400 (from 4)      |
+        +-------+------+-------+------------------+-------------------+
+        | false | 4    | A     | 40               | 400               |
+        +-------+------+-------+------------------+-------------------+
+        | true  | 5    | A     | 40 (from 4)      | 600 (from 6)      |
+        +-------+------+-------+------------------+-------------------+
+        | false | 6    | A     | 60               | 600               |
+        +-------+------+-------+------------------+-------------------+
+        | false | 1    | B     | -10              | -100              |
+        +-------+------+-------+------------------+-------------------+
+        | true  | 2    | B     | -10 (from 1)     | -300 (from 3)     |
+        +-------+------+-------+------------------+-------------------+
+        | false | 3    | B     | -30              | -300              |
+        +-------+------+-------+------------------+-------------------+
+    )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(TimeSeriesImputerTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Times",
+          "Tensor of timestamps in seconds since epoch [R] where R is a number of rows.",
+          "T1")
+      .Input(
+          2,
+          "Keys",
+          "Composite keys tensor of shape [R][K]. R is the same as Input(1)",
+          "T2")
+      .Input(
+          3,
+          "Data",
+          "It is a data tensor of shape [R][C] where R - rows and C - columns. R must be the same with Input(1)",
+          "T2")
+      .Output(
+          0,
+          "Added",
+          "Tensor of boolean with a shape of [IR]. Contains a boolean for each row in the result where true represents added row.",
+          "T3")
+      .Output(
+          1,
+          "ImputedTimes",
+          "This is a tensor of timestamps in seconds since epoch of shape [IR], where IR is the number of output rows.",
+          "T1")
+      .Output(
+          2,
+          "ImputedKeys",
+          "Contains keys along with the imputed keys. Tensor of shape [IR][K].",
+          "T2")
+      .Output(
+          3,
+          "ImputedData",
+          "Tensor of shape [IR][C] where IR is the number of rows in the output."
+          "C is the number of columns.",
+          "T2")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "T1",
+          {"tensor(int64)"},
+          "Represents number of seconds since epoch")
+      .TypeConstraint(
+          "T2",
+          {"tensor(string)"},
+          "Output data")
+      .TypeConstraint(
+          "T3",
+          {"tensor(bool)"},
+          "Boolean Tensor")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_BOOL, 0);
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT64, 1);
+            // Number of output rows is not known
+            ONNX_NAMESPACE::TensorShapeProto shape_0_1;
+            shape_0_1.add_dim();
+            ONNX_NAMESPACE::updateOutputShape(ctx, 0, shape_0_1);
+            ONNX_NAMESPACE::updateOutputShape(ctx, 1, shape_0_1);
+
+            // Keys
+            propagateElemTypeFromInputToOutput(ctx, 2, 2);
+            // Keys shape
+            if (hasInputShape(ctx, 2)) {
+              const auto& input2_shape = getInputShape(ctx, 2);
+              if (input2_shape.dim_size() != 2) {
+                fail_shape_inference("Expecting keys to have 2 dimensions");
+              }
+              ONNX_NAMESPACE::TensorShapeProto shape;
+              shape.add_dim();
+              *shape.add_dim() = input2_shape.dim(1);
+              ONNX_NAMESPACE::updateOutputShape(ctx, 2, shape);
+            }
+
+            // Data shape
+            propagateElemTypeFromInputToOutput(ctx, 3, 3);
+            if (hasInputShape(ctx, 3)) {
+              const auto& input3_shape = getInputShape(ctx, 3);
+              if (input3_shape.dim_size() != 2) {
+                fail_shape_inference("Expecting data to have 2 dimensions");
+              }
+              ONNX_NAMESPACE::TensorShapeProto shape;
+              shape.add_dim();
+              *shape.add_dim() = input3_shape.dim(1);
+              ONNX_NAMESPACE::updateOutputShape(ctx, 3, shape);
+            }
           });
 }
 
diff --git a/onnxruntime/featurizers_ops/cpu/time_series_imputer_transformer.cc b/onnxruntime/featurizers_ops/cpu/time_series_imputer_transformer.cc
new file mode 100644
index 0000000000000..ea8f4755eab37
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/time_series_imputer_transformer.cc
@@ -0,0 +1,244 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/op_kernel.h"
+
+#include <cstdlib>
+#include <limits>
+
+#include "Featurizers/TimeSeriesImputerFeaturizer.h"
+#include "Archive.h"
+
+namespace ft = Microsoft::Featurizer::Featurizers;
+
+namespace onnxruntime {
+namespace featurizers {
+
+namespace timeseries_imputer_details {
+
+inline std::chrono::system_clock::time_point ToTimePoint(int64_t secs) {
+  return std::chrono::system_clock::from_time_t(secs);
+}
+
+inline int64_t ToSecs(const std::chrono::system_clock::time_point& tp) {
+  using namespace std::chrono;
+  return duration_cast<seconds>(tp.time_since_epoch()).count();
+}
+
+template <typename T>
+struct ToString {
+  std::string operator()(T val) const {
+    return std::to_string(val);
+  }
+};
+
+template <>
+struct ToString<std::string> {
+  const std::string& operator()(const std::string& val) const {
+    return val;
+  }
+};
+
+template <typename T>
+struct ToStringOptional {
+  nonstd::optional<std::string> operator()(T val) const {
+    nonstd::optional<std::string> result;
+    if (std::isnan(val)) {
+      return result;
+    }
+    result = std::to_string(val);
+    return result;
+  }
+};
+
+template <>
+struct ToStringOptional<std::string> {
+  nonstd::optional<std::string> operator()(std::string val) const {
+    return (val.empty()) ? nonstd::optional<std::string>() : nonstd::optional<std::string>(std::move(val));
+  }
+};
+
+template <typename T>
+struct FromString;
+
+template <>
+struct FromString<std::string> {
+  const std::string& operator()(const std::string& val) const {
+    return val;
+  }
+};
+
+template <>
+struct FromString<float> {
+  float operator()(const std::string& val) const {
+    char* str_end = nullptr;
+    const char* str = val.c_str();
+    float result = std::strtof(str, &str_end);
+    if (str == str_end) {
+      ORT_THROW("Resulting key string is not convertible to float: ", val);
+    }
+    return result;
+  }
+};
+
+template <>
+struct FromString<double> {
+  double operator()(const std::string& val) const {
+    const char* str = val.c_str();
+    char* str_end = nullptr;
+    double result = std::strtod(str, &str_end);
+    if (str == str_end) {
+      ORT_THROW("Resulting key string is not convertible to double: ", val);
+    }
+    return result;
+  }
+};
+template <typename T>
+struct FromStringOptional {
+  T operator()(const nonstd::optional<std::string>& val) const {
+    if (val.has_value()) {
+      return FromString<T>()(*val);
+    }
+    return std::numeric_limits<T>::quiet_NaN();
+  }
+};
+
+template <>
+struct FromStringOptional<std::string> {
+  std::string operator()(const nonstd::optional<std::string>& val) const {
+    if (val.has_value()) {
+      return *val;
+    }
+    return std::string();
+  }
+};
+}  // namespace timeseries_imputer_details
+
+template <typename T>
+struct TimeSeriesImputerTransformerImpl {
+  void operator()(OpKernelContext* ctx, int64_t rows) {
+    const auto& state = *ctx->Input<Tensor>(0);
+    const uint8_t* const state_data = state.template Data<uint8_t>();
+
+    const auto& times = *ctx->Input<Tensor>(1);
+    const auto& keys = *ctx->Input<Tensor>(2);
+    const auto& data = *ctx->Input<Tensor>(3);
+
+    const int64_t keys_per_row = keys.Shape()[1];
+    const int64_t columns = data.Shape()[1];
+
+    using namespace timeseries_imputer_details;
+
+    using OutputType = std::tuple<bool, std::chrono::system_clock::time_point,
+                                  std::vector<std::string>, std::vector<nonstd::optional<std::string>>>;
+    std::vector<OutputType> output_rows;
+    std::function<void(OutputType)> callback_fn;
+    callback_fn = [&output_rows](OutputType value) -> void {
+      output_rows.emplace_back(std::move(value));
+    };
+
+    Microsoft::Featurizer::Archive archive(state_data, state.Shape().Size());
+    ft::Components::TimeSeriesImputerEstimator::Transformer transformer(archive);
+
+    const int64_t* times_data = times.template Data<int64_t>();
+    const T* const keys_data = keys.template Data<T>();
+    const T* const data_data = data.template Data<T>();
+
+    // for each row get timestamp, get all keys, get all data and feed it
+    for (int64_t row = 0; row < rows; ++row) {
+      const T* const key_row_data = keys_data + (row * keys_per_row);
+      const T* const keys_row_end = key_row_data + keys_per_row;
+      std::vector<std::string> str_keys;
+      std::transform(key_row_data, keys_row_end, std::back_inserter(str_keys),
+                     ToString<T>());
+
+      std::vector<nonstd::optional<std::string>> str_data;
+      const T* const data_row = data_data + (row * columns);
+      const T* const data_row_end = data_row + columns;
+      std::transform(data_row, data_row_end, std::back_inserter(str_data),
+                     ToStringOptional<T>());
+
+      auto tuple_row = std::make_tuple(ToTimePoint(*times_data), std::move(str_keys), std::move(str_data));
+
+      transformer.execute(tuple_row, callback_fn);
+      ++times_data;
+    }
+
+    transformer.flush(callback_fn);
+
+    // Compute output shapes now
+    // Number of outputs is the number of rows,
+    int64_t output_rows_num = static_cast<int64_t>(output_rows.size());
+    TensorShape rows_shape({output_rows_num});
+    TensorShape keys_shape({output_rows_num, keys_per_row});
+    TensorShape data_shape({output_rows_num, columns});
+
+    auto* added_output = ctx->Output(0, rows_shape)->template MutableData<bool>();
+    auto* time_output = ctx->Output(1, rows_shape)->template MutableData<int64_t>();
+    auto* keys_output = ctx->Output(2, keys_shape)->template MutableData<T>();
+    auto* data_output = ctx->Output(3, data_shape)->template MutableData<T>();
+
+    for (const auto& out : output_rows) {
+      *added_output++ = std::get<0>(out);
+      *time_output++ = ToSecs(std::get<1>(out));
+      const auto& imputed_keys = std::get<2>(out);
+      ORT_ENFORCE(static_cast<int64_t>(imputed_keys.size()) == keys_per_row,
+                  "resulting number of keys: ", imputed_keys.size(), " expected: ", keys_per_row);
+      const auto& imputed_data = std::get<3>(out);
+      ORT_ENFORCE(static_cast<int64_t>(imputed_data.size()) == columns,
+                  "resulting number of columns: ", imputed_data.size(), " expected: ", columns);
+      keys_output = std::transform(imputed_keys.cbegin(), imputed_keys.cend(), keys_output,
+                                   FromString<T>());
+      data_output = std::transform(imputed_data.cbegin(), imputed_data.cend(), data_output,
+                                   FromStringOptional<T>());
+    }
+  }
+};
+
+class TimeSeriesImputerTransformer final : public OpKernel {
+ public:
+  explicit TimeSeriesImputerTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  static Status CheckBatches(int64_t rows, const TensorShape& shape) {
+    if (shape.NumDimensions() == 2) {
+      ORT_RETURN_IF_NOT(rows == shape[0], "Number of rows does not match");
+    } else {
+      return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT, "Expect shape of [R][C]");
+    }
+    return Status::OK();
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    const auto& times = *ctx->Input<Tensor>(1);
+    const auto& times_shape = times.Shape();
+    ORT_RETURN_IF_NOT(times_shape.NumDimensions() == 1, "Times must have shape [B][R] or [R]");
+    int64_t rows = times_shape[0];
+
+    const auto& keys = *ctx->Input<Tensor>(2);
+    ORT_RETURN_IF_ERROR(CheckBatches(rows, keys.Shape()));
+    const auto& data = *ctx->Input<Tensor>(3);
+    ORT_RETURN_IF_ERROR(CheckBatches(rows, data.Shape()));
+
+    auto data_type = data.GetElementType();
+    ORT_RETURN_IF_NOT(keys.GetElementType() == data_type, "Keys and data must have the same datatype");
+
+    TimeSeriesImputerTransformerImpl<std::string>()(ctx, rows);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    TimeSeriesImputerTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("T1", DataTypeImpl::GetTensorType<int64_t>())
+        .TypeConstraint("T2", DataTypeImpl::GetTensorType<std::string>()),
+    TimeSeriesImputerTransformer);
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc b/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
index 5f246c894b2b9..2fa7168880743 100644
--- a/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
+++ b/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
@@ -19,6 +19,7 @@ class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomai
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MissingDummiesTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, RobustScalarTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, StringTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, TimeSeriesImputerTransformer);
 
 Status RegisterCpuMSFeaturizersKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
@@ -31,6 +32,7 @@ Status RegisterCpuMSFeaturizersKernels(KernelRegistry& kernel_registry) {
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MissingDummiesTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, RobustScalarTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, StringTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, TimeSeriesImputerTransformer)>,
   };
 
   for (auto& function_table_entry : function_table) {
diff --git a/onnxruntime/test/featurizers_ops/time_series_imputer_transformer_test.cc b/onnxruntime/test/featurizers_ops/time_series_imputer_transformer_test.cc
new file mode 100644
index 0000000000000..d8a351d248088
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/time_series_imputer_transformer_test.cc
@@ -0,0 +1,196 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Featurizers/TimeSeriesImputerFeaturizer.h"
+#include "Featurizers/TestHelpers.h"
+#include "Archive.h"
+
+namespace NS = Microsoft::Featurizer;
+
+namespace onnxruntime {
+namespace test {
+
+inline std::chrono::system_clock::time_point GetTimePoint(std::chrono::system_clock::time_point tp, int unitsToAdd, std::string = "days") {
+  return tp + std::chrono::minutes(unitsToAdd * (60 * 24));
+}
+
+inline int64_t GetTimeSecs(std::chrono::system_clock::time_point tp) {
+  using namespace std::chrono;
+  return time_point_cast<seconds>(tp).time_since_epoch().count();
+}
+
+using InputType = std::tuple<
+    std::chrono::system_clock::time_point,
+    std::vector<std::string>,
+    std::vector<nonstd::optional<std::string>>>;
+
+using TransformedType = std::vector<
+    std::tuple<
+        bool,
+        std::chrono::system_clock::time_point,
+        std::vector<std::string>,
+        std::vector<nonstd::optional<std::string>>>>;
+
+std::vector<uint8_t> GetStream(const std::vector<std::vector<InputType>>& trainingBatches,
+                               const std::vector<NS::TypeId>& colsToImputeDataTypes,
+                               bool supressError, NS::Featurizers::Components::TimeSeriesImputeStrategy tsImputeStrategy) {
+  using TSImputerEstimator = NS::Featurizers::TimeSeriesImputerEstimator;
+
+  NS::AnnotationMapsPtr const pAllColumnAnnotations(NS::CreateTestAnnotationMapsPtr(1));
+  TSImputerEstimator estimator(pAllColumnAnnotations, colsToImputeDataTypes, supressError, tsImputeStrategy);
+
+  NS::TestHelpers::Train<TSImputerEstimator, InputType>(estimator, trainingBatches);
+  TSImputerEstimator::TransformerUniquePtr pTransformer(estimator.create_transformer());
+
+  NS::Archive ar;
+  pTransformer->save(ar);
+  return ar.commit();
+}
+
+static void AddInputs(OpTester& test, const std::vector<std::vector<InputType>>& trainingBatches,
+                      const std::vector<InputType>& inferenceBatches, const std::vector<NS::TypeId>& colsToImputeDataTypes,
+                      bool supressError, NS::Featurizers::Components::TimeSeriesImputeStrategy tsImputeStrategy) {
+  auto stream = GetStream(
+      trainingBatches,
+      colsToImputeDataTypes,
+      supressError,
+      tsImputeStrategy);
+
+  auto dim = static_cast<int64_t>(stream.size());
+  test.AddInput<uint8_t>("State", {dim}, stream);
+
+  std::vector<int64_t> times;
+  std::vector<std::string> keys;
+  std::vector<std::string> data;
+
+  using namespace std::chrono;
+  for (const auto& infb : inferenceBatches) {
+    times.push_back(time_point_cast<seconds>(std::get<0>(infb)).time_since_epoch().count());
+    keys.insert(keys.end(), std::get<1>(infb).cbegin(), std::get<1>(infb).cend());
+    std::transform(std::get<2>(infb).cbegin(), std::get<2>(infb).cend(), std::back_inserter(data),
+                   [](const nonstd::optional<std::string>& opt) -> std::string {
+                     if (opt.has_value()) return *opt;
+                     return std::string();
+                   });
+  }
+
+  // Should have equal amount of keys per row
+  ASSERT_TRUE(keys.size() % times.size() == 0);
+  ASSERT_TRUE(data.size() % times.size() == 0);
+  test.AddInput<int64_t>("Times", {static_cast<int64_t>(times.size())}, times);
+  test.AddInput<std::string>("Keys", {static_cast<int64_t>(times.size()), static_cast<int64_t>(keys.size() / times.size())}, keys);
+  test.AddInput<std::string>("Data", {static_cast<int64_t>(times.size()), static_cast<int64_t>(data.size() / times.size())}, data);
+}
+
+void AddOutputs(OpTester& test, const std::initializer_list<bool>& added, const std::initializer_list<std::chrono::system_clock::time_point>& times,
+                const std::vector<std::string>& keys, const std::vector<std::string>& data) {
+  ASSERT_TRUE(keys.size() % times.size() == 0);
+  ASSERT_TRUE(data.size() % times.size() == 0);
+
+  std::vector<int64_t> times_int64;
+  std::transform(times.begin(), times.end(), std::back_inserter(times_int64), GetTimeSecs);
+
+  test.AddOutput<bool>("Added", {static_cast<int64_t>(added.size())}, added);
+  test.AddOutput<int64_t>("ImputedTimes", {static_cast<int64_t>(times.size())}, times_int64);
+  test.AddOutput<std::string>("ImputedKeys", {static_cast<int64_t>(times.size()), static_cast<int64_t>(keys.size() / times.size())}, keys);
+  test.AddOutput<std::string>("ImputedData", {static_cast<int64_t>(times.size()), static_cast<int64_t>(data.size() / times.size())}, data);
+}
+
+TEST(FeaturizersTests, RowImputation_1_grain_no_gaps) {
+  std::chrono::system_clock::time_point now = std::chrono::system_clock::now();
+  auto tp_0 = GetTimePoint(now, 0);
+  auto tp_1 = GetTimePoint(now, 1);
+  auto tp_2 = GetTimePoint(now, 2);
+  auto tuple_1 = std::make_tuple(tp_0, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{"14.5", "18"});
+  auto tuple_2 = std::make_tuple(tp_1, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_3 = std::make_tuple(tp_2, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{"15.0", nonstd::optional<std::string>{}});
+
+  std::vector<InputType> inferenceBatches = {tuple_1,
+                                             tuple_2,
+                                             tuple_3};
+
+  OpTester test("TimeSeriesImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+
+  AddInputs(test, {inferenceBatches}, inferenceBatches,
+            {NS::TypeId::Float64, NS::TypeId::Float64}, false, NS::Featurizers::Components::TimeSeriesImputeStrategy::Forward);
+  AddOutputs(test, {false, false, false}, {tp_0, tp_1, tp_2},
+             {"a", "a", "a"}, {"14.5", "18", "14.5", "12", "15.0", "12"});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, RowImputation_1_grain_2_gaps) {
+  std::chrono::system_clock::time_point now = std::chrono::system_clock::now();
+  auto tp_0 = GetTimePoint(now, 0);
+  auto tp_1 = GetTimePoint(now, 1);
+  auto tp_2 = GetTimePoint(now, 2);
+  auto tp_3 = GetTimePoint(now, 3);
+
+  auto tuple_0 = std::make_tuple(tp_0, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{"14.5", "18"});
+  auto tuple_1 = std::make_tuple(tp_1, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_3 = std::make_tuple(tp_3, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "15.0"});
+
+  OpTester test("TimeSeriesImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  AddInputs(test, {{tuple_0, tuple_1}}, {tuple_0, tuple_3},
+            {NS::TypeId::Float64, NS::TypeId::Float64}, false, NS::Featurizers::Components::TimeSeriesImputeStrategy::Forward);
+
+  AddOutputs(test, {false, true, true, false}, {tp_0, tp_1, tp_2, tp_3},
+             {"a", "a", "a", "a"}, {"14.5", "18", "14.5", "18", "14.5", "18", "14.5", "15.0"});
+  test.Run();
+}
+
+TEST(FeaturizersTests, RowImputation_2_grains_no_gaps_input_interleaved) {
+  std::chrono::system_clock::time_point now = std::chrono::system_clock::now();
+  auto tp_0 = GetTimePoint(now, 0);
+  auto tp_1 = GetTimePoint(now, 1);
+  auto tp_5 = GetTimePoint(now, 5);
+  auto tp_6 = GetTimePoint(now, 6);
+
+  auto tuple_0 = std::make_tuple(tp_0, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{"14.5", "18"});
+  auto tuple_5 = std::make_tuple(tp_5, std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{"14.5", "18"});
+  auto tuple_5_inf = std::make_tuple(GetTimePoint(now, 5), std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{"114.5", "118"});
+  auto tuple_1 = std::make_tuple(tp_1, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_6 = std::make_tuple(tp_6, std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_6_inf = std::make_tuple(GetTimePoint(now, 6), std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "112"});
+
+  OpTester test("TimeSeriesImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  AddInputs(test, {{tuple_0, tuple_5, tuple_1, tuple_6}}, {tuple_0, tuple_5_inf, tuple_1, tuple_6_inf},
+            {NS::TypeId::Float64, NS::TypeId::Float64}, false, NS::Featurizers::Components::TimeSeriesImputeStrategy::Forward);
+
+  AddOutputs(test, {false, false, false, false}, {tp_0, tp_5, tp_1, tp_6},
+             {"a", "b", "a", "b"}, {"14.5", "18", "114.5", "118", "14.5", "12", "114.5", "112"});
+  test.Run();
+}
+
+TEST(FeaturizersTests, RowImputation_2_grains_1_gap_input_interleaved) {
+  std::chrono::system_clock::time_point now = std::chrono::system_clock::now();
+  auto tp_0 = GetTimePoint(now, 0);
+  auto tp_1 = GetTimePoint(now, 1);
+  auto tp_2 = GetTimePoint(now, 2);
+  auto tp_5 = GetTimePoint(now, 5);
+  auto tp_6 = GetTimePoint(now, 6);
+  auto tp_7 = GetTimePoint(now, 7);
+
+  auto tuple_0 = std::make_tuple(tp_0, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{"14.5", "18"});
+  auto tuple_2 = std::make_tuple(GetTimePoint(now, 2), std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_5 = std::make_tuple(tp_5, std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{"14.5", "18"});
+  auto tuple_5_inf = std::make_tuple(tp_5, std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{"114.5", "118"});
+  auto tuple_1 = std::make_tuple(tp_1, std::vector<std::string>{"a"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_6 = std::make_tuple(tp_6, std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "12"});
+  auto tuple_7 = std::make_tuple(GetTimePoint(now, 7), std::vector<std::string>{"b"}, std::vector<nonstd::optional<std::string>>{nonstd::optional<std::string>{}, "112"});
+
+  OpTester test("TimeSeriesImputerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  AddInputs(test, {{tuple_0, tuple_5, tuple_1, tuple_6}}, {tuple_0, tuple_5_inf, tuple_2, tuple_7},
+            {NS::TypeId::Float64, NS::TypeId::Float64}, false, NS::Featurizers::Components::TimeSeriesImputeStrategy::Forward);
+
+  AddOutputs(test, {false, false, true, false, true, false}, {tp_0, tp_5, tp_1, tp_2, tp_6, tp_7},
+             {"a", "b", "a", "a", "b", "b"}, {"14.5", "18", "114.5", "118", "14.5", "18", "14.5", "12", "114.5", "118", "114.5", "112"});
+
+  test.Run();
+}
+
+}  // namespace test
+}  // namespace onnxruntime

From c4f6db7796dffef06d392948fdfe7cc7a977b3b3 Mon Sep 17 00:00:00 2001
From: stevenlix <38092805+stevenlix@users.noreply.github.com>
Date: Fri, 10 Jan 2020 14:07:40 -0800
Subject: [PATCH 43/75] Fix memory leak in TRT (#2815)

* fix memory leak issue

* revert EP_FAIL on enueueV2
---
 .../tensorrt/tensorrt_execution_provider.cc      | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider.cc b/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider.cc
index 028a34e09d2e0..1a56b8229e2c1 100644
--- a/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider.cc
+++ b/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider.cc
@@ -676,7 +676,7 @@ common::Status TensorrtExecutionProvider::Compile(const std::vector<onnxruntime:
       // Update shape ranges
       bool dimension_update = false;
       auto trt_builder = trt_state->builder;
-      auto trt_profile = trt_builder->createOptimizationProfile();
+      nvinfer1::IOptimizationProfile* trt_profile = nullptr;
       for (int i = 0, end = num_binding_inputs; i < end; ++i) {
         // TODO: check if getInput indexing is same with binding index
         auto input = trt_state->network->getInput(i);
@@ -714,6 +714,9 @@ common::Status TensorrtExecutionProvider::Compile(const std::vector<onnxruntime:
           }
 
           if (dimension_update) {
+            if (trt_profile == nullptr) {
+              trt_profile = trt_builder->createOptimizationProfile();
+            }
             if (engine.isShapeBinding(i)) {
               std::vector<int32_t> shapes_min(nb_dims), shapes_opt(nb_dims), shapes_max(nb_dims);
               for (int j = 0, end = nb_dims; j < end; ++j) {
@@ -733,7 +736,7 @@ common::Status TensorrtExecutionProvider::Compile(const std::vector<onnxruntime:
         }
 
         // TensorRT6 requires optimization profile to be defined for all inputs if any input dimension is symbolic
-        if (dynamic_shape) {
+        if (dimension_update && dynamic_shape) {
           trt_profile->setDimensions(input->getName(), nvinfer1::OptProfileSelector::kMIN, dims_min);
           trt_profile->setDimensions(input->getName(), nvinfer1::OptProfileSelector::kOPT, dims_opt);
           trt_profile->setDimensions(input->getName(), nvinfer1::OptProfileSelector::kMAX, dims_max);
@@ -746,10 +749,13 @@ common::Status TensorrtExecutionProvider::Compile(const std::vector<onnxruntime:
         auto trt_config = unique_pointer<nvinfer1::IBuilderConfig>(trt_builder->createBuilderConfig());
         trt_config->addOptimizationProfile(trt_profile);
         trt_state->engine = trt_builder->buildEngineWithConfig(*trt_state->network, *trt_config);
-        ORT_ENFORCE(trt_state->engine != nullptr);
-
+        if (trt_state->engine == nullptr) {
+          return ORT_MAKE_STATUS(ONNXRUNTIME, EP_FAIL, "TensorRT EP Failed to Build Engine.");
+        }
         trt_state->context = trt_state->engine->createExecutionContext();
-        ORT_ENFORCE(trt_state->context != nullptr);
+        if (trt_state->context == nullptr) {
+          return ORT_MAKE_STATUS(ONNXRUNTIME, EP_FAIL, "TensorRT EP Failed to Create Context.");
+        }
         trt_context = trt_state->context;
       }
 

From e37cdbed74e237cdd467fdd5433eae9c5da85d2f Mon Sep 17 00:00:00 2001
From: Dmitri Smirnov <dmitrism@microsoft.com>
Date: Fri, 10 Jan 2020 13:28:57 -0800
Subject: [PATCH 44/75] Add manifest missing comma

---
 cgmanifest.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/cgmanifest.json b/cgmanifest.json
index 62ad0a3031625..331efe0d32d80 100644
--- a/cgmanifest.json
+++ b/cgmanifest.json
@@ -446,7 +446,7 @@
             },
             "type": "git"
          }
-      }
+      },
       {
          "component": {
             "git": {

From c4e4abce73965ff7819d3183712ce40e2ca13cb4 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Fri, 10 Jan 2020 22:17:17 -0800
Subject: [PATCH 45/75] Run static code analyzer on most of our code (#2817)

---
 .../github/azure-pipelines/templates/compliance.yml      | 9 +++++++++
 .../github/azure-pipelines/templates/win-ci-2019.yml     | 1 -
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/tools/ci_build/github/azure-pipelines/templates/compliance.yml b/tools/ci_build/github/azure-pipelines/templates/compliance.yml
index f96857856bc4a..87e3c21bfa6e7 100644
--- a/tools/ci_build/github/azure-pipelines/templates/compliance.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/compliance.yml
@@ -11,6 +11,15 @@ steps:
     arguments: 'analyze $(Build.BinariesDirectory)\RelWithDebInfo\*.dll  --recurse --verbose'
   continueOnError: true
 
+- task: DeleteFiles@1
+  displayName: 'Delete files from $(Build.BinariesDirectory)\RelWithDebInfo'
+  inputs:
+    SourceFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
+    Contents: |
+     **/*.obj
+     **/*.pdb
+     **/*.dll
+      
 - task: securedevelopmentteam.vss-secure-development-tools.build-task-prefast.SDLNativeRules@2
   displayName: 'Run the PREfast SDL Native Rules for MSBuild'
   continueOnError: true
diff --git a/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml b/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml
index d4ae4da1b7ec1..b32e6958cf8d7 100644
--- a/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/win-ci-2019.yml
@@ -84,7 +84,6 @@ jobs:
         logProjectEvents: true
         workingFolder: '$(Build.BinariesDirectory)\RelWithDebInfo'
         createLogFile: true
-        clean: true
 
     # Build RelWithDebInfo -- this variable required to build C#
     - script: |

From 8d017f57013485ea88ceff76a3c7ab71ce1e7d14 Mon Sep 17 00:00:00 2001
From: Ryan Lai <ryalai96@gmail.com>
Date: Mon, 13 Jan 2020 09:22:54 -0800
Subject: [PATCH 46/75] Scneario Test : Build Google Test and Taef Test based
 on preprocessor definition (#2809)

* Add winml macro wrappers on top of google test macros

* change test methods to disabled

* Add custom winml macros for both taef and google tests

* PR comments
---
 cmake/winml_unittests.cmake                   |    2 +
 winml/test/common/SqueezeNetValidator.cpp     |   48 +-
 winml/test/common/SqueezeNetValidator.h       |    2 +-
 winml/test/common/googleTestMacros.h          |   63 +
 winml/test/common/protobufHelpers.cpp         |   39 +-
 winml/test/common/runtimeParameters.h         |    2 +-
 winml/test/common/std.h                       |   34 +-
 winml/test/common/taefTestMacros.h            |   59 +
 winml/test/common/test.h                      |   10 +
 winml/test/common/testPch.h                   |    2 +
 winml/test/scenario/cppwinrt/CustomNullOp.h   |    8 +-
 .../cppwinrt/CustomOperatorProvider.h         |    2 +-
 winml/test/scenario/cppwinrt/CustomOps.cpp    |  127 +-
 winml/test/scenario/cppwinrt/CustomOps.h      |   19 +
 .../cppwinrt/scenariotestscppwinrt.cpp        | 2497 ++++++++---------
 .../scenario/cppwinrt/scenariotestscppwinrt.h |   85 +
 16 files changed, 1561 insertions(+), 1438 deletions(-)
 create mode 100644 winml/test/common/googleTestMacros.h
 create mode 100644 winml/test/common/taefTestMacros.h
 create mode 100644 winml/test/common/test.h
 create mode 100644 winml/test/scenario/cppwinrt/CustomOps.h
 create mode 100644 winml/test/scenario/cppwinrt/scenariotestscppwinrt.h

diff --git a/cmake/winml_unittests.cmake b/cmake/winml_unittests.cmake
index 18e6f7cdf85ff..95661e85d8bc2 100644
--- a/cmake/winml_unittests.cmake
+++ b/cmake/winml_unittests.cmake
@@ -60,6 +60,7 @@ add_dependencies(winml_test_common
   winml_api
   winml_dll
 )
+target_compile_definitions(winml_test_common PRIVATE BUILD_GOOGLE_TEST)
 set_winml_target_properties(winml_test_common)
 
 file(GLOB winml_test_api_src CONFIGURE_DEPENDS "${WINML_TEST_SRC_DIR}/api/*.cpp")
@@ -82,6 +83,7 @@ add_winml_test(
   LIBS winml_test_common ${winml_test_scenario_libs}
 )
 target_precompiled_header(winml_test_scenario testPch.h)
+target_compile_definitions(winml_test_scenario PRIVATE BUILD_GOOGLE_TEST)
 set_target_properties(winml_test_scenario PROPERTIES LINK_FLAGS
   "/DELAYLOAD:d2d1.dll /DELAYLOAD:d3d11.dll /DELAYLOAD:dxgi.dll"
 )
diff --git a/winml/test/common/SqueezeNetValidator.cpp b/winml/test/common/SqueezeNetValidator.cpp
index e3fc0978ad086..2eacc456cd017 100644
--- a/winml/test/common/SqueezeNetValidator.cpp
+++ b/winml/test/common/SqueezeNetValidator.cpp
@@ -2,12 +2,11 @@
 #include "protobufHelpers.h"
 #include "fileHelpers.h"
 #include "core/common/common.h"
-#include <gtest/gtest.h>
 #include <winrt/Windows.Media.h>
 #include <winrt/Windows.Graphics.Imaging.h>
 #include <winrt/Windows.Storage.h>
 #include <winrt/Windows.Storage.Streams.h>
-
+#include <iostream>
 // using namespace winrt::Windows::Foundation;
 using namespace winrt::Windows::AI::MachineLearning;
 using namespace winrt::Windows::Foundation::Collections;
@@ -35,12 +34,12 @@ static void BindImage(
 
     if (bindAsInspectable)
     {
-        EXPECT_NO_THROW(binding.Bind(name, frame));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, frame));
     }
     else
     {
         auto imagetensor = ImageFeatureValue::CreateFromVideoFrame(frame);
-        EXPECT_NO_THROW(binding.Bind(name, imagetensor));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, imagetensor));
     }
 }
 
@@ -50,15 +49,15 @@ static void BindTensor(
     ITensor inputTensor,
     bool bindAsInspectable = false)
 {
-    EXPECT_TRUE(inputTensor != nullptr);
+    WINML_EXPECT_TRUE(inputTensor != nullptr);
 
     if (bindAsInspectable)
     {
-        EXPECT_NO_THROW(binding.Bind(name, inputTensor.as<TensorFloat>().GetAsVectorView()));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, inputTensor.as<TensorFloat>().GetAsVectorView()));
     }
     else
     {
-        EXPECT_NO_THROW(binding.Bind(name, inputTensor));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, inputTensor));
     }
 }
 
@@ -75,11 +74,11 @@ ITensor BindOutput(
     {
     case OutputBindingStrategy::Bound:
         outputTensor = T::Create(shape);
-        EXPECT_NO_THROW(binding.Bind(name, outputTensor));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, outputTensor));
         break;
     case OutputBindingStrategy::Empty:
         outputTensor = T::Create();
-        EXPECT_NO_THROW(binding.Bind(name, outputTensor));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, outputTensor));
         break;
     case OutputBindingStrategy::Unbound:
         __fallthrough;
@@ -104,7 +103,7 @@ ImageFeatureValue BindImageOutput(
         SoftwareBitmap bitmap(BitmapPixelFormat::Bgra8, 720, 720);
         VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(bitmap);
         outputTensor = ImageFeatureValue::CreateFromVideoFrame(frame);
-        EXPECT_NO_THROW(binding.Bind(name, outputTensor));
+        WINML_EXPECT_NO_THROW(binding.Bind(name, outputTensor));
         break;
     }
     case OutputBindingStrategy::Unbound:
@@ -136,10 +135,10 @@ void ModelValidator::FnsCandy16(
 
     // WinML model creation
     LearningModel model = nullptr;
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
+    WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
 
     LearningModelSession modelSession = nullptr;
-    EXPECT_NO_THROW(modelSession = LearningModelSession(model, LearningModelDevice(deviceKind)));
+    WINML_EXPECT_NO_THROW(modelSession = LearningModelSession(model, LearningModelDevice(deviceKind)));
 
     LearningModelBinding modelBinding(modelSession);
     auto fullImagePath = modulePath + inputDataImageFileName;
@@ -147,7 +146,7 @@ void ModelValidator::FnsCandy16(
 
     // create the tensor for the actual output
     auto output = model.OutputFeatures().First().Current();
-    EXPECT_TRUE(output.Kind() == LearningModelFeatureKind::Tensor);
+    WINML_EXPECT_TRUE(output.Kind() == LearningModelFeatureKind::Tensor);
 
     auto shape = winrt::single_threaded_vector(std::vector<int64_t> {1, 1});
     auto outputTensor = BindImageOutput(outputBindingStrategy, modelBinding, outputDataBindingName);
@@ -155,7 +154,7 @@ void ModelValidator::FnsCandy16(
     // Evaluate the model
     std::cout << "Calling EvaluateSync on instance" << instance << "\n";
     LearningModelEvaluationResult result = nullptr;
-    EXPECT_NO_THROW(result = modelSession.Evaluate(modelBinding, {}));
+    WINML_EXPECT_NO_THROW(result = modelSession.Evaluate(modelBinding, {}));
 
     // Get results
     if (outputBindingStrategy == OutputBindingStrategy::Unbound)
@@ -167,7 +166,7 @@ void ModelValidator::FnsCandy16(
     }
     else
     {
-        EXPECT_EQ(result.Outputs().Lookup(outputDataBindingName), outputTensor);
+        WINML_EXPECT_EQUAL(result.Outputs().Lookup(outputDataBindingName), outputTensor);
 
         auto softwareBitmap = outputTensor.VideoFrame().SoftwareBitmap();
 
@@ -203,10 +202,10 @@ void ModelValidator::SqueezeNet(
     
     // WinML model creation
     LearningModel model = nullptr;
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
+    WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
     
     LearningModelSession modelSession = nullptr;
-    EXPECT_NO_THROW(modelSession = LearningModelSession(model, LearningModelDevice(deviceKind)));
+    WINML_EXPECT_NO_THROW(modelSession = LearningModelSession(model, LearningModelDevice(deviceKind)));
 
     LearningModelBinding modelBinding(modelSession);
 
@@ -224,11 +223,11 @@ void ModelValidator::SqueezeNet(
 
     // load up the expected output
     auto expectedResultsTensor = ProtobufHelpers::LoadTensorFromProtobufFile(outputFileName, false);
-    EXPECT_TRUE(expectedResultsTensor != nullptr);
+    WINML_EXPECT_TRUE(expectedResultsTensor != nullptr);
 
     // create the tensor for the actual output
     auto output = model.OutputFeatures().First().Current();
-    EXPECT_TRUE(output.Kind() == LearningModelFeatureKind::Tensor);
+    WINML_EXPECT_TRUE(output.Kind() == LearningModelFeatureKind::Tensor);
 
     auto outputTensor = BindOutput<TensorFloat>(
         outputBindingStrategy, modelBinding, outputDataBindingName, expectedResultsTensor.Shape());
@@ -236,7 +235,7 @@ void ModelValidator::SqueezeNet(
     // Evaluate the model
     std::cout << "Calling EvaluateSync on instance" << instance << "\n";
     LearningModelEvaluationResult result = nullptr;
-    EXPECT_NO_THROW(result = modelSession.Evaluate(modelBinding, {}));
+    WINML_EXPECT_NO_THROW(result = modelSession.Evaluate(modelBinding, {}));
 
     // Get results
     if (outputBindingStrategy == OutputBindingStrategy::Unbound)
@@ -247,21 +246,22 @@ void ModelValidator::SqueezeNet(
     }
     else
     {
-        EXPECT_EQ(result.Outputs().Lookup(outputDataBindingName), outputTensor);
+        WINML_EXPECT_EQUAL(result.Outputs().Lookup(outputDataBindingName), outputTensor);
     }
 
     auto outDataExpected = expectedResultsTensor.as<TensorFloat>().GetAsVectorView();
     auto outDataActual = outputTensor.as<TensorFloat>().GetAsVectorView();
 
-    EXPECT_TRUE(outDataActual.Size() == outDataExpected.Size());
+    WINML_EXPECT_TRUE(outDataActual.Size() == outDataExpected.Size());
     for (uint32_t i = 0; i < outDataActual.Size(); i++)
     {
         float delta = std::abs(outDataActual.GetAt(i) - outDataExpected.GetAt(i));
         if (delta > dataTolerance)
         {
-            ADD_FAILURE() << "EXPECTED: " << outDataExpected.GetAt(i) << " , ACTUAL: " << outDataActual.GetAt(i)
+          std::stringstream ss;
+          ss << "EXPECTED: " << outDataExpected.GetAt(i) << " , ACTUAL: " << outDataActual.GetAt(i)
                 << "instance " << instance << ", element " << i;
-
+          WINML_LOG_ERROR(ss.str().c_str());
         }
     }
 }
diff --git a/winml/test/common/SqueezeNetValidator.h b/winml/test/common/SqueezeNetValidator.h
index 68b44ddb2dd3b..613df41853fbe 100644
--- a/winml/test/common/SqueezeNetValidator.h
+++ b/winml/test/common/SqueezeNetValidator.h
@@ -6,7 +6,7 @@
 
 #pragma once
 
-#include <std.h>
+#include "std.h"
 
 enum OutputBindingStrategy { Bound, Unbound, Empty };
 
diff --git a/winml/test/common/googleTestMacros.h b/winml/test/common/googleTestMacros.h
new file mode 100644
index 0000000000000..55fb2f7f740fe
--- /dev/null
+++ b/winml/test/common/googleTestMacros.h
@@ -0,0 +1,63 @@
+#include <gtest/gtest.h>
+#include "runtimeParameters.h"
+
+#define TEST_GROUP_BEGIN(group_name)
+#define TEST_GROUP_END()
+
+#define WINML_TEST(group_name, test_name) \
+  TEST_F(group_name, test_name) {         \
+    getapi().test_name();                 \
+  }
+
+#define WINML_TEST_CLASS_BEGIN_NO_SETUP(test_class_name) \
+  class test_class_name : public ::testing::Test {       \
+  };
+
+#define WINML_TEST_CLASS_BEGIN_WITH_SETUP(test_class_name, setup_method) \
+  class test_class_name : public ::testing::Test {                       \
+   protected:                                                            \
+    void SetUp() override {                                              \
+      getapi().setup_method();                                           \
+    }                                                                    \
+  };
+
+#define WINML_TEST_CLASS_END()
+
+// For old versions of gtest without GTEST_SKIP, stream the message and return success instead
+#ifndef GTEST_SKIP
+#define GTEST_SKIP_(message) \
+  return GTEST_MESSAGE_(message, ::testing::TestPartResult::kSuccess)
+#define GTEST_SKIP GTEST_SKIP_("")
+#endif
+
+#define WINML_SKIP_TEST(message) \
+  GTEST_SKIP() << message;
+
+#define WINML_EXPECT_NO_THROW(statement) EXPECT_NO_THROW(statement)
+#define WINML_EXPECT_TRUE(statement) EXPECT_TRUE(statement)
+#define WINML_EXPECT_EQUAL(val1, val2) EXPECT_EQ(val1, val2)
+#define WINML_EXPECT_NOT_EQUAL(val1, val2) EXPECT_NE(val1, val2)
+
+#define WINML_LOG_ERROR(message) \
+  ADD_FAILURE() << message
+
+#define WINML_EXPECT_HRESULT_SUCCEEDED(hresult_expression) EXPECT_HRESULT_SUCCEEDED(hresult_expression)
+#define WINML_EXPECT_HRESULT_FAILED(hresult_expression) EXPECT_HRESULT_FAILED(hresult_expression)
+#define WINML_EXPECT_THROW_SPECIFIC(statement, exception, condition) EXPECT_THROW_SPECIFIC(statement, exception, condition)
+
+#ifndef USE_DML
+#define GPUTEST \
+  WINML_SKIP_TEST("GPU tests disabled because this is a WinML only build (no DML)")
+#else
+#define GPUTEST                                                                         \
+  if (auto noGpuTests = RuntimeParameters::Parameters.find("noGPUtests");               \
+      noGpuTests != RuntimeParameters::Parameters.end() && noGpuTests->second != "0") { \
+    WINML_SKIP_TEST("GPU tests disabled");                                              \
+  }
+#endif
+
+#define SKIP_EDGECORE                                                                   \
+  if (auto isEdgeCore = RuntimeParameters::Parameters.find("EdgeCore");                 \
+      isEdgeCore != RuntimeParameters::Parameters.end() && isEdgeCore->second != "0") { \
+    WINML_SKIP_TEST("Test can't be run in EdgeCore");                                   \
+  }
\ No newline at end of file
diff --git a/winml/test/common/protobufHelpers.cpp b/winml/test/common/protobufHelpers.cpp
index a5bc12a70e4a2..0f3c22fb1c36b 100644
--- a/winml/test/common/protobufHelpers.cpp
+++ b/winml/test/common/protobufHelpers.cpp
@@ -1,10 +1,11 @@
-#define _SILENCE_ALL_CXX17_DEPRECATION_WARNINGS
+#ifndef _SILENCE_ALL_CXX17_DEPRECATION_WARNINGS
+#define _SILENCE_ALL_CXX17_DEPRECATION_WARNINGS
+#endif
 
 // LotusRT
 #include "core/framework/allocatormgr.h"
 #include "core/common/logging/logging.h"
 #include "core/common/logging/sinks/clog_sink.h"
-
 #include "protobufHelpers.h"
 
 #pragma warning(push)
@@ -12,7 +13,6 @@
 #include "onnx/onnx-ml.pb.h"
 #pragma warning(pop)
 
-#include <gtest/gtest.h>
 #include <fstream>
 
 #include "winrt/Windows.Storage.Streams.h"
@@ -66,33 +66,35 @@ bool LoadTensorFromPb(onnx::TensorProto& tensor, std::wstring filePath) {
 
 template <typename DataType>
 std::vector<DataType> GetTypeSpecificDataFromTensorProto(
-  onnx::TensorProto /*tensorProto*/){
+    onnx::TensorProto /*tensorProto*/) {
   static_assert(false, "UNDEFINED! TensorProto methods aren't templated, so add a new template specialization.");
 }
 template <>
 std::vector<float> GetTypeSpecificDataFromTensorProto(
-  onnx::TensorProto tensorProto){
+    onnx::TensorProto tensorProto) {
   return std::vector<float>(std::begin(tensorProto.float_data()), std::end(tensorProto.float_data()));
 }
 template <>
 std::vector<int32_t> GetTypeSpecificDataFromTensorProto(
-  onnx::TensorProto tensorProto){
+    onnx::TensorProto tensorProto) {
   return std::vector<int32_t>(std::begin(tensorProto.int32_data()), std::end(tensorProto.int32_data()));
 }
 template <>
 std::vector<int64_t> GetTypeSpecificDataFromTensorProto(
-  onnx::TensorProto tensorProto){
+    onnx::TensorProto tensorProto) {
   return std::vector<int64_t>(std::begin(tensorProto.int64_data()), std::end(tensorProto.int64_data()));
 }
 
 template <typename DataType>
 std::vector<DataType> GetTensorDataFromTensorProto(
-	onnx::TensorProto tensorProto,
-	uint64_t elementCount) {
+    onnx::TensorProto tensorProto,
+    uint64_t elementCount) {
   if (tensorProto.has_raw_data()) {
     std::vector<DataType> tensorData;
     auto& values = tensorProto.raw_data();
-    EXPECT_EQ(elementCount, values.size() / sizeof(DataType)) << L"TensorProto elementcount should match raw data buffer size in elements.";
+    if (elementCount != values.size() / sizeof(DataType)) {
+        WINML_LOG_ERROR("TensorProto element count should match raw data buffer size in elements.");
+    }
 
     tensorData = std::vector<DataType>(elementCount);
     memcpy(tensorData.data(), values.data(), values.size());
@@ -105,7 +107,7 @@ std::vector<DataType> GetTensorDataFromTensorProto(
 static std::vector<winrt::hstring> GetTensorStringDataFromTensorProto(
     onnx::TensorProto tensorProto,
     uint64_t elementCount) {
-  EXPECT_EQ(tensorProto.string_data_size(), elementCount);
+  WINML_EXPECT_EQUAL(tensorProto.string_data_size(), elementCount);
   auto& values = tensorProto.string_data();
   auto returnVector = std::vector<winrt::hstring>(elementCount);
   std::transform(std::begin(values), std::end(values), std::begin(returnVector),
@@ -131,15 +133,15 @@ ITensor ProtobufHelpers::LoadTensorFromProtobufFile(
     }
     switch (tensorProto.data_type()) {
       case (onnx::TensorProto::DataType::TensorProto_DataType_FLOAT):
-       return TensorFloat::CreateFromIterable(tensorShape, GetTensorDataFromTensorProto<float>(tensorProto, elementCount));
+        return TensorFloat::CreateFromIterable(tensorShape, GetTensorDataFromTensorProto<float>(tensorProto, elementCount));
       case (onnx::TensorProto::DataType::TensorProto_DataType_INT32):
-       return TensorInt32Bit::CreateFromIterable(tensorShape, GetTensorDataFromTensorProto<int32_t>(tensorProto, elementCount));
+        return TensorInt32Bit::CreateFromIterable(tensorShape, GetTensorDataFromTensorProto<int32_t>(tensorProto, elementCount));
       case (onnx::TensorProto::DataType::TensorProto_DataType_INT64):
-       return TensorInt64Bit::CreateFromIterable(tensorShape, GetTensorDataFromTensorProto<int64_t>(tensorProto, elementCount));
+        return TensorInt64Bit::CreateFromIterable(tensorShape, GetTensorDataFromTensorProto<int64_t>(tensorProto, elementCount));
       case (onnx::TensorProto::DataType::TensorProto_DataType_STRING):
         return TensorString::CreateFromIterable(tensorShape, GetTensorStringDataFromTensorProto(tensorProto, elementCount));
       default:
-        ADD_FAILURE() << L"Tensor type for creating tensor from protobuf file not supported.";
+        WINML_LOG_ERROR("Tensor type for creating tensor from protobuf file not supported.");
         break;
     }
   }
@@ -152,7 +154,7 @@ TensorFloat16Bit ProtobufHelpers::LoadTensorFloat16FromProtobufFile(
   onnx::TensorProto tensorProto;
   if (LoadTensorFromPb(tensorProto, filePath)) {
     if (tensorProto.has_data_type()) {
-      EXPECT_EQ(onnx::TensorProto::DataType::TensorProto_DataType_FLOAT16, tensorProto.data_type());
+      WINML_EXPECT_EQUAL(onnx::TensorProto::DataType::TensorProto_DataType_FLOAT16, tensorProto.data_type());
     } else {
       std::cerr << "Loading unknown TensorProto datatype as TensorFloat16Bit.\n";
     }
@@ -166,7 +168,10 @@ TensorFloat16Bit ProtobufHelpers::LoadTensorFloat16FromProtobufFile(
     uint32_t sizeInBytes;
     spTensorValueNative->GetBuffer(reinterpret_cast<BYTE**>(&data), &sizeInBytes);
 
-    EXPECT_TRUE(tensorProto.has_raw_data()) << L"Float16 tensor proto buffers are expected to contain raw data.";
+    if (!tensorProto.has_raw_data())
+    {
+      WINML_LOG_ERROR("Float16 tensor proto buffers are expected to contain raw data.");
+    }
 
     auto& raw_data = tensorProto.raw_data();
     auto buff = raw_data.c_str();
diff --git a/winml/test/common/runtimeParameters.h b/winml/test/common/runtimeParameters.h
index 6b0edc7a9cc8b..a98e6c5e283a0 100644
--- a/winml/test/common/runtimeParameters.h
+++ b/winml/test/common/runtimeParameters.h
@@ -6,4 +6,4 @@ namespace RuntimeParameters
 {
     // Runtime parameters passed through CLI arguments
     extern std::unordered_map<std::string, std::string> Parameters;
-}
+}
\ No newline at end of file
diff --git a/winml/test/common/std.h b/winml/test/common/std.h
index a2ea803ab70a9..162915abd02fd 100644
--- a/winml/test/common/std.h
+++ b/winml/test/common/std.h
@@ -19,7 +19,7 @@
 #include <utility>
 #include <vector>
 
-#include <gtest/gtest.h>
+#include "test.h"
 
 // IUnknown must be declared before winrt/base.h is included to light up support for native COM
 // interfaces with C++/WinRT types (e.g. winrt::com_ptr<ITensorNative>).
@@ -31,8 +31,6 @@
 // WinML
 #include "Windows.AI.MachineLearning.Native.h"
 
-#include "runtimeParameters.h"
-
 #define EXPECT_THROW_SPECIFIC(statement, exception, condition)  \
     EXPECT_THROW(                                               \
         try {                                                   \
@@ -43,35 +41,7 @@
         }                                                       \
     , exception);
 
-// For old versions of gtest without GTEST_SKIP, stream the message and return success instead
-#ifndef GTEST_SKIP
-#define GTEST_SKIP_(message) \
-    return GTEST_MESSAGE_(message, ::testing::TestPartResult::kSuccess)
-#define GTEST_SKIP GTEST_SKIP_("")
-#endif
-
 #ifndef INSTANTIATE_TEST_SUITE_P
 // Use the old name, removed in newer versions of googletest
 #define INSTANTIATE_TEST_SUITE_P INSTANTIATE_TEST_CASE_P
-#endif
-
-
-#ifndef USE_DML
-#define GPUTEST \
-    GTEST_SKIP() << "GPU tests disabled because this is a WinML only build (no DML)";
-#else
-#define GPUTEST \
-    if (auto noGpuTests = RuntimeParameters::Parameters.find("noGPUtests");             \
-        noGpuTests != RuntimeParameters::Parameters.end() && noGpuTests->second != "0") \
-    {                                                                                   \
-        GTEST_SKIP() << "GPU tests disabled";                                           \
-    }
-#endif 
-    
-
-#define SKIP_EDGECORE \
-    if (auto isEdgeCore = RuntimeParameters::Parameters.find("EdgeCore");               \
-        isEdgeCore != RuntimeParameters::Parameters.end() && isEdgeCore->second != "0") \
-    {                                                                                   \
-        GTEST_SKIP() << "Test can't be run in EdgeCore";                                \
-    }
+#endif
\ No newline at end of file
diff --git a/winml/test/common/taefTestMacros.h b/winml/test/common/taefTestMacros.h
new file mode 100644
index 0000000000000..7c7636b7a4dda
--- /dev/null
+++ b/winml/test/common/taefTestMacros.h
@@ -0,0 +1,59 @@
+#include "WexTestClass.h"
+
+using namespace WEX::Logging;
+using namespace WEX::Common;
+using namespace WEX::TestExecution;
+
+#define WINML_EXPECT_NO_THROW(statement) VERIFY_NO_THROW(statement)
+
+#define WINML_TEST_CLASS_BEGIN_WITH_SETUP(test_class_name, setup_method) \
+  class test_class_name {                                                \
+    TEST_CLASS(test_class_name);                                         \
+    TEST_CLASS_SETUP(TestClassSetup) {                                   \
+      getapi().setup_method();                                           \
+      return true;                                                       \
+    }
+
+#define WINML_TEST_CLASS_END() \
+  }                            \
+  ;
+
+#define WINML_TEST(group_name, test_name) \
+  TEST_METHOD(test_name) {                \
+    getapi().test_name();                 \
+  }
+
+#define WINML_SKIP_TEST(message)                                                               \
+  Log::Result(TestResults::Skipped,                                                            \
+              std::wstring_convert<std::codecvt_utf8<wchar_t>>().from_bytes(message).c_str()); \
+  return;
+
+#define WINML_EXPECT_NO_THROW(statement) VERIFY_NO_THROW(statement)
+#define WINML_EXPECT_TRUE(statement) VERIFY_IS_TRUE(statement)
+#define WINML_EXPECT_EQUAL(val1, val2) VERIFY_ARE_EQUAL(val1, val2)
+#define WINML_EXPECT_NOT_EQUAL(val1, val2) VERIFY_ARE_NOT_EQUAL(val1, val2)
+#define WINML_LOG_ERROR(message) \
+  VERIFY_FAIL(std::wstring_convert<std::codecvt_utf8<wchar_t>>().from_bytes(message).c_str())
+
+#define WINML_EXPECT_HRESULT_SUCCEEDED(hresult_expression) VERIFY_SUCCEEDED(hresult_expression)
+#define WINML_EXPECT_THROW_SPECIFIC(statement, exception, condition) VERIFY_THROWS_SPECIFIC(statement, exception, condition)
+#define WINML_EXPECT_HRESULT_FAILED(hresult_expression) VERIFY_FAILED(hresult_expression)
+
+#ifndef USE_DML
+#define GPUTEST \
+  WINML_SKIP_TEST("GPU tests disabled because this is a WinML only build (no DML)")
+#else
+#define GPUTEST                                                                             \
+  bool noGPUTests;                                                                          \
+  if (SUCCEEDED(RuntimeParameters::TryGetValue(L"noGPUtests", noGPUTests)) && noGPUTests) { \
+    WINML_SKIP_TEST("This test is disabled by the noGPUTests runtime parameter.");          \
+    return;                                                                                 \
+  }
+#endif
+
+#define SKIP_EDGECORE                                                                       \
+  bool edgeCoreRun;                                                                         \
+  if (SUCCEEDED(RuntimeParameters::TryGetValue(L"EdgeCore", edgeCoreRun)) && edgeCoreRun) { \
+    WINML_SKIP_TEST("This test is disabled by the EdgeCore runtime parameter.");            \
+    return;                                                                                 \
+  }
\ No newline at end of file
diff --git a/winml/test/common/test.h b/winml/test/common/test.h
new file mode 100644
index 0000000000000..31739f672503f
--- /dev/null
+++ b/winml/test/common/test.h
@@ -0,0 +1,10 @@
+using VoidTest = void (*)();
+using SetupTest = VoidTest;
+
+#ifdef BUILD_GOOGLE_TEST
+#include "googleTestMacros.h"
+#else
+#ifdef BUILD_TAEF_TEST
+#include "taefTestMacros.h"
+#endif
+#endif
\ No newline at end of file
diff --git a/winml/test/common/testPch.h b/winml/test/common/testPch.h
index 952f0bbccf6a1..4bc156fa0c1c2 100644
--- a/winml/test/common/testPch.h
+++ b/winml/test/common/testPch.h
@@ -3,7 +3,9 @@
 //  Copyright (c) Microsoft Corporation. All rights reserved.
 //
 //-----------------------------------------------------------------------------
+#ifndef _SILENCE_ALL_CXX17_DEPRECATION_WARNINGS
 #define _SILENCE_ALL_CXX17_DEPRECATION_WARNINGS
+#endif
 #include "std.h"
 
 #include <wrl/client.h>
diff --git a/winml/test/scenario/cppwinrt/CustomNullOp.h b/winml/test/scenario/cppwinrt/CustomNullOp.h
index d7dfdf75ac488..7d2e5efa899f4 100644
--- a/winml/test/scenario/cppwinrt/CustomNullOp.h
+++ b/winml/test/scenario/cppwinrt/CustomNullOp.h
@@ -4,14 +4,14 @@
 
 #pragma once
 
-#include <gtest/gtest.h>
+#include "test.h"
 
 template <typename T>
 struct NullShapeInferrer : winrt::implements<NullShapeInferrer<T>, IMLOperatorShapeInferrer>
 {
     STDMETHOD(InferOutputShapes)(IMLOperatorShapeInferenceContext* context) noexcept
     {
-        EXPECT_NO_THROW(OperatorHelper::ShapeInferenceFunction<T>(context));
+        WINML_EXPECT_NO_THROW(OperatorHelper::ShapeInferenceFunction<T>(context));
         return S_OK;
     }
 };
@@ -23,7 +23,7 @@ struct NullOperator : winrt::implements<NullOperator, IMLOperatorKernel>
     STDMETHOD(Compute)(IMLOperatorKernelContext* context)
     {
         winrt::com_ptr<IMLOperatorTensor> outputTensor;
-        EXPECT_HRESULT_SUCCEEDED(context->GetOutputTensor(0, outputTensor.put()));
+        WINML_EXPECT_HRESULT_SUCCEEDED(context->GetOutputTensor(0, outputTensor.put()));
 
         ++(*m_callCount);
         return S_OK;
@@ -92,7 +92,7 @@ struct NullOperatorFactory : winrt::implements<NullOperatorFactory, IMLOperatorK
 
         auto factory = winrt::make<NullOperatorFactory>(callCount);
 
-        EXPECT_HRESULT_SUCCEEDED(registry->RegisterOperatorKernel(
+        WINML_EXPECT_HRESULT_SUCCEEDED(registry->RegisterOperatorKernel(
             &kernelDescription,
             factory.get(),
             shapeInferrer.get()
diff --git a/winml/test/scenario/cppwinrt/CustomOperatorProvider.h b/winml/test/scenario/cppwinrt/CustomOperatorProvider.h
index 215cd4c4f29b8..698eeeb396095 100644
--- a/winml/test/scenario/cppwinrt/CustomOperatorProvider.h
+++ b/winml/test/scenario/cppwinrt/CustomOperatorProvider.h
@@ -54,4 +54,4 @@ struct CustomOperatorProvider :
         m_registry.copy_to(ppOperatorRegistry);
         return S_OK;
     }
-};
+};
\ No newline at end of file
diff --git a/winml/test/scenario/cppwinrt/CustomOps.cpp b/winml/test/scenario/cppwinrt/CustomOps.cpp
index a9ab916fbc1bd..eb47843b95de6 100644
--- a/winml/test/scenario/cppwinrt/CustomOps.cpp
+++ b/winml/test/scenario/cppwinrt/CustomOps.cpp
@@ -12,7 +12,7 @@
 #include <MemoryBuffer.h>
 #include <gsl/gsl>
 #include "CustomOperatorProvider.h"
-#include "runtimeParameters.h"
+#include "CustomOps.h"
 
 // For custom operator and shape inferencing support
 #include "core/providers/dml/DmlExecutionProvider/inc/MLOperatorAuthor.h"
@@ -31,26 +31,19 @@ using namespace winrt::Windows::Graphics::Imaging;
 using namespace winrt::Windows::Storage;
 using namespace winrt::Windows::Storage::Streams;
 
-class CustomOpsScenarioTest : public ::testing::Test
+static void CustomOpsScenarioTestSetup()
 {
-protected:
-    CustomOpsScenarioTest() {
-        init_apartment();
-    }
-};
+  init_apartment();
+}
 
-class CustomOpsScenarioGpuTest : public CustomOpsScenarioTest
+static void CustomOpsScenarioGpuTestSetup()
 {
-protected:
-    void SetUp() override
-    {
-        GPUTEST
-    }
-};
+  init_apartment();
+  GPUTEST;
+}
 
 // Tests that the execution provider correctly fuses operators together when custom ops are involved.
-TEST_F(CustomOpsScenarioGpuTest, CustomOperatorFusion)
-{
+static void CustomOperatorFusion() {
     constexpr const wchar_t* c_modelFilename = L"squeezenet_tensor_input.onnx";
 
     // This particular model has 25 Conv ops and 25 Relu ops, all of which are eligible for fusion so we expect them
@@ -96,7 +89,7 @@ TEST_F(CustomOpsScenarioGpuTest, CustomOperatorFusion)
         {
             using namespace OperatorHelper;
 
-            EXPECT_HRESULT_SUCCEEDED(MLCreateOperatorRegistry(m_registry.put()));
+            WINML_EXPECT_HRESULT_SUCCEEDED(MLCreateOperatorRegistry(m_registry.put()));
 
 #pragma push_macro("REGISTER_KERNEL")
 #define REGISTER_KERNEL(_name, _domain, _opSet, _shapeInferrer, _callCount) \
@@ -143,14 +136,14 @@ TEST_F(CustomOpsScenarioGpuTest, CustomOperatorFusion)
     auto provider = customOperatorProvider.as<ILearningModelOperatorProvider>();
 
     LearningModelDevice device = nullptr;
-    EXPECT_NO_THROW(device = LearningModelDevice(LearningModelDeviceKind::DirectX));
+    WINML_EXPECT_NO_THROW(device = LearningModelDevice(LearningModelDeviceKind::DirectX));
     std::wstring fullPath = FileHelpers::GetModulePath() + c_modelFilename;
     auto model = LearningModel::LoadFromFilePath(fullPath, provider);
 
     auto featureValue = FileHelpers::LoadImageFeatureValue(L"227x227.png");
 
     LearningModelSession session = nullptr;
-    EXPECT_NO_THROW(session = LearningModelSession(model, device));
+    WINML_EXPECT_NO_THROW(session = LearningModelSession(model, device));
     LearningModelBinding modelBinding(session);
 
     modelBinding.Bind(L"data", featureValue);
@@ -159,15 +152,15 @@ TEST_F(CustomOpsScenarioGpuTest, CustomOperatorFusion)
     const auto& callCounts = customOperatorProvider.as<CallbackOperatorProvider>()->GetCallCounts();
 
     // Verify that the correct number of each operator was seen (i.e. that none were dropped / incorrectly fused)
-    EXPECT_EQ(c_expectedConvOps, callCounts.conv);
-    EXPECT_EQ(c_expectedReluOps, callCounts.relu);
-    EXPECT_EQ(c_expectedFusedConvOps, callCounts.fusedConv);
-    EXPECT_EQ(c_expectedGemmOps, callCounts.gemm);
-    EXPECT_EQ(c_expectedSigmoidOps, callCounts.sigmoid);
-    EXPECT_EQ(c_expectedFusedGemmOps, callCounts.fusedGemm);
-    EXPECT_EQ(c_expectedBatchNormOps, callCounts.batchNorm);
-    EXPECT_EQ(c_expectedMaxPoolOps, callCounts.maxPool);
-    EXPECT_EQ(c_expectedConcatOps, callCounts.concat);
+    WINML_EXPECT_EQUAL(c_expectedConvOps, callCounts.conv);
+    WINML_EXPECT_EQUAL(c_expectedReluOps, callCounts.relu);
+    WINML_EXPECT_EQUAL(c_expectedFusedConvOps, callCounts.fusedConv);
+    WINML_EXPECT_EQUAL(c_expectedGemmOps, callCounts.gemm);
+    WINML_EXPECT_EQUAL(c_expectedSigmoidOps, callCounts.sigmoid);
+    WINML_EXPECT_EQUAL(c_expectedFusedGemmOps, callCounts.fusedGemm);
+    WINML_EXPECT_EQUAL(c_expectedBatchNormOps, callCounts.batchNorm);
+    WINML_EXPECT_EQUAL(c_expectedMaxPoolOps, callCounts.maxPool);
+    WINML_EXPECT_EQUAL(c_expectedConcatOps, callCounts.concat);
 }
 
 struct LocalCustomOperatorProvider :
@@ -178,7 +171,7 @@ struct LocalCustomOperatorProvider :
 {
     LocalCustomOperatorProvider()
     {
-        EXPECT_HRESULT_SUCCEEDED(MLCreateOperatorRegistry(m_registry.put()));
+        WINML_EXPECT_HRESULT_SUCCEEDED(MLCreateOperatorRegistry(m_registry.put()));
     }
 
     STDMETHOD(GetRegistry)(IMLOperatorRegistry** ppOperatorRegistry)
@@ -205,20 +198,20 @@ struct LocalCustomOperatorProvider :
 void VerifyTestAttributes(const MLOperatorAttributes& attrs)
 {
     std::string strAttr = attrs.GetAttribute("DefaultedNonRequiredString");
-    EXPECT_EQ(strAttr, "1");
+    WINML_EXPECT_EQUAL(strAttr, "1");
 
     std::vector<std::string> strArrayAttr = attrs.GetAttributeVector("DefaultedNonRequiredStringArray");
     std::vector<std::string> expected = std::vector<std::string>({ "1", "2" });
     for (size_t i = 0; i < expected.size(); ++i)
     {
-        EXPECT_EQ(strArrayAttr[i], expected[i]);
+        WINML_EXPECT_EQUAL(strArrayAttr[i], expected[i]);
     }
 
-    EXPECT_EQ(1, attrs.GetAttribute<int64_t>("DefaultedNonRequiredInt"));
-    EXPECT_EQ(1.0f, attrs.GetAttribute<float>("DefaultedNonRequiredFloat"));
+    WINML_EXPECT_EQUAL(1, attrs.GetAttribute<int64_t>("DefaultedNonRequiredInt"));
+    WINML_EXPECT_EQUAL(1.0f, attrs.GetAttribute<float>("DefaultedNonRequiredFloat"));
 
-    EXPECT_EQ(std::vector<int64_t>({ 1, 2 }), attrs.GetAttributeVector<int64_t>("DefaultedNonRequiredIntArray"));
-    EXPECT_EQ(std::vector<float>({ 1.0f, 2.0f }), attrs.GetAttributeVector<float>("DefaultedNonRequiredFloatArray"));
+    WINML_EXPECT_EQUAL(std::vector<int64_t>({ 1, 2 }), attrs.GetAttributeVector<int64_t>("DefaultedNonRequiredIntArray"));
+    WINML_EXPECT_EQUAL(std::vector<float>({ 1.0f, 2.0f }), attrs.GetAttributeVector<float>("DefaultedNonRequiredFloatArray"));
 }
 
 // Foo kernel which is doing Add and optionally truncates its output
@@ -241,14 +234,14 @@ class FooKernel
         if (!Truncate)
         {
             com_ptr<IMLOperatorTensorShapeDescription> shapeInfo;
-            EXPECT_EQ(info.GetInterface()->HasTensorShapeDescription(), false);
-            EXPECT_HRESULT_FAILED(info.GetInterface()->GetTensorShapeDescription(shapeInfo.put()));
+            WINML_EXPECT_EQUAL(info.GetInterface()->HasTensorShapeDescription(), false);
+            WINML_EXPECT_HRESULT_FAILED(info.GetInterface()->GetTensorShapeDescription(shapeInfo.put()));
         }
         else
         {
             com_ptr<IMLOperatorTensorShapeDescription> shapeInfo;
-            EXPECT_EQ(info.GetInterface()->HasTensorShapeDescription(), true);
-            EXPECT_EQ(info.GetInterface()->GetTensorShapeDescription(shapeInfo.put()), S_OK);
+            WINML_EXPECT_EQUAL(info.GetInterface()->HasTensorShapeDescription(), true);
+            WINML_EXPECT_EQUAL(info.GetInterface()->GetTensorShapeDescription(shapeInfo.put()), S_OK);
         }
     }
 
@@ -271,7 +264,7 @@ class FooKernel
         if (!Truncate)
         {
             com_ptr<IMLOperatorTensor> tensor;
-            EXPECT_HRESULT_FAILED(context.GetInterface()->GetOutputTensor(0, tensor.put()));
+            WINML_EXPECT_HRESULT_FAILED(context.GetInterface()->GetOutputTensor(0, tensor.put()));
         }
         else
         {
@@ -308,7 +301,7 @@ void CreateTruncatedABIFooKernel(IMLOperatorKernelCreationContext* kernelInfo, I
 }
 
 // Test using a foo kernel which is doing Add, but register it as "Mul".
-TEST_F(CustomOpsScenarioTest, CustomKernelWithBuiltInSchema)
+static void CustomKernelWithBuiltInSchema()
 {
     // Create the registry
     auto operatorProvider = winrt::make<LocalCustomOperatorProvider>();
@@ -337,7 +330,7 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithBuiltInSchema)
     };
 
     Microsoft::WRL::ComPtr<MLOperatorKernelFactory> factory = wil::MakeOrThrow<MLOperatorKernelFactory>(CreateABIFooKernel<false>);
-    EXPECT_HRESULT_SUCCEEDED(registry->RegisterOperatorKernel(&kernelDesc, factory.Get(), nullptr));
+    WINML_EXPECT_HRESULT_SUCCEEDED(registry->RegisterOperatorKernel(&kernelDesc, factory.Get(), nullptr));
 
     // Prepare inputs
     std::vector<int64_t> dimsX = { 3, 2 };
@@ -361,23 +354,23 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithBuiltInSchema)
     bindings.Bind(winrt::hstring(L"X"), inputTensor);
 
     auto outputValue = TensorFloat::Create();
-    EXPECT_NO_THROW(bindings.Bind(L"Y", outputValue));
+    WINML_EXPECT_NO_THROW(bindings.Bind(L"Y", outputValue));
 
     // Evaluate the model
     hstring correlationId;
-    EXPECT_NO_THROW(session.Evaluate(bindings, correlationId));
+    WINML_EXPECT_NO_THROW(session.Evaluate(bindings, correlationId));
 
     // Check the result shape
-    EXPECT_EQ(expectedDimsY.size(), outputValue.Shape().Size());
+    WINML_EXPECT_EQUAL(expectedDimsY.size(), outputValue.Shape().Size());
     for (uint32_t j = 0; j < outputValue.Shape().Size(); j++)
     {
-        EXPECT_EQ(expectedDimsY.at(j), outputValue.Shape().GetAt(j));
+        WINML_EXPECT_EQUAL(expectedDimsY.at(j), outputValue.Shape().GetAt(j));
     }
 
     // Check the results
     auto buffer = outputValue.GetAsVectorView();
-    EXPECT_TRUE(buffer != nullptr);
-    EXPECT_TRUE(std::equal(expectedValuesY.cbegin(), expectedValuesY.cend(), begin(buffer)));
+    WINML_EXPECT_TRUE(buffer != nullptr);
+    WINML_EXPECT_TRUE(std::equal(expectedValuesY.cbegin(), expectedValuesY.cend(), begin(buffer)));
 
     // Release the model before operatorProvider goes out of scope
     model = nullptr;
@@ -404,7 +397,7 @@ class MLOperatorShapeInferrerFromFunc : public Microsoft::WRL::RuntimeClass<
 };
 
 // Test using a custom kernel and schema, while verifying attribute defaults, type mapping, and inference methods
-TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
+static void CustomKernelWithCustomSchema()
 {
     // Test cases
     struct
@@ -591,7 +584,7 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
         // Register the schema
         MLOperatorSetId opsetId = { "", 7 };
         MLOperatorSchemaDescription* opSchemaDescs = &schemaDesc;
-        EXPECT_EQ(S_OK, registry->RegisterOperatorSetSchema(
+        WINML_EXPECT_EQUAL(S_OK, registry->RegisterOperatorSetSchema(
             &opsetId,
             1,
             &opSchemaDescs,
@@ -608,7 +601,7 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
 
             MLOperatorSetId id = { "", 9 };
             MLOperatorSchemaDescription* schemaDescs = &futureSchemaDesc;
-            EXPECT_EQ(S_OK, registry->RegisterOperatorSetSchema(
+            WINML_EXPECT_EQUAL(S_OK, registry->RegisterOperatorSetSchema(
                 &id,
                 7,
                 &schemaDescs,
@@ -624,7 +617,7 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
 
             MLOperatorSetId id = { "otherDomain", 7 };
             MLOperatorSchemaDescription* schemaDescs = &otherSchemaDesc;
-            EXPECT_EQ(S_OK, registry->RegisterOperatorSetSchema(
+            WINML_EXPECT_EQUAL(S_OK, registry->RegisterOperatorSetSchema(
                 &id,
                 1,
                 &schemaDescs,
@@ -661,12 +654,12 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
             kernelDesc.options = MLOperatorKernelOptions::AllowDynamicInputShapes;
             Microsoft::WRL::ComPtr<MLOperatorKernelFactory> factory = wil::MakeOrThrow<MLOperatorKernelFactory>(CreateABIFooKernel<true>);
 
-            EXPECT_EQ(S_OK, registry->RegisterOperatorKernel(&kernelDesc, factory.Get(), nullptr));
+            WINML_EXPECT_EQUAL(S_OK, registry->RegisterOperatorKernel(&kernelDesc, factory.Get(), nullptr));
         }
         else
         {
             Microsoft::WRL::ComPtr<MLOperatorKernelFactory> factory = wil::MakeOrThrow<MLOperatorKernelFactory>(CreateTruncatedABIFooKernel);
-            EXPECT_EQ(S_OK, registry->RegisterOperatorKernel(
+            WINML_EXPECT_EQUAL(S_OK, registry->RegisterOperatorKernel(
                 &kernelDesc,
                 factory.Get(),
                 testCases[caseIndex].useShapeInferenceInKernel ? shapeInferrer.Get() : nullptr
@@ -699,23 +692,23 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
         bindings.Bind(winrt::hstring(L"X"), inputTensor);
 
         auto outputValue = TensorFloat::Create();
-        EXPECT_NO_THROW(bindings.Bind(L"Y", outputValue));
+        WINML_EXPECT_NO_THROW(bindings.Bind(L"Y", outputValue));
 
         // Evaluate the model
         hstring correlationId;
-        EXPECT_NO_THROW(session.Evaluate(bindings, correlationId));
+        WINML_EXPECT_NO_THROW(session.Evaluate(bindings, correlationId));
 
         // Verify the result shape
-        EXPECT_EQ(expectedDimsY.size(), outputValue.Shape().Size());
+        WINML_EXPECT_EQUAL(expectedDimsY.size(), outputValue.Shape().Size());
         for (uint32_t j = 0; j < outputValue.Shape().Size(); j++)
         {
-            EXPECT_EQ(expectedDimsY.at(j), outputValue.Shape().GetAt(j));
+            WINML_EXPECT_EQUAL(expectedDimsY.at(j), outputValue.Shape().GetAt(j));
         }
 
         // Verify the result values
         auto buffer = outputValue.GetAsVectorView();
-        EXPECT_TRUE(buffer != nullptr);
-        EXPECT_TRUE(std::equal(expectedValuesY.cbegin(), expectedValuesY.cend(), begin(buffer)));
+        WINML_EXPECT_TRUE(buffer != nullptr);
+        WINML_EXPECT_TRUE(std::equal(expectedValuesY.cbegin(), expectedValuesY.cend(), begin(buffer)));
 
         // Release the model before operatorProvider goes out of scope
         model = nullptr;
@@ -724,7 +717,19 @@ TEST_F(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
         {
             // Check that the shape inference context is closed and safely fails
             MLOperatorEdgeDescription edgeDesc;
-            EXPECT_EQ(E_INVALIDARG, shapeInferenceContext->GetInputEdgeDescription(0, &edgeDesc));
+            WINML_EXPECT_EQUAL(E_INVALIDARG, shapeInferenceContext->GetInputEdgeDescription(0, &edgeDesc));
         }
     }
 }
+
+const CustomOpsTestApi& getapi() {
+  static constexpr CustomOpsTestApi api = 
+      {
+          CustomOpsScenarioTestSetup,
+          CustomOpsScenarioGpuTestSetup,
+          CustomOperatorFusion,
+          CustomKernelWithBuiltInSchema,
+          CustomKernelWithCustomSchema
+      };
+  return api;
+}
\ No newline at end of file
diff --git a/winml/test/scenario/cppwinrt/CustomOps.h b/winml/test/scenario/cppwinrt/CustomOps.h
new file mode 100644
index 0000000000000..df48bd51af990
--- /dev/null
+++ b/winml/test/scenario/cppwinrt/CustomOps.h
@@ -0,0 +1,19 @@
+#include "test.h"
+struct CustomOpsTestApi
+{
+    SetupTest CustomOpsScenarioTestSetup;
+    SetupTest CustomOpsScenarioGpuTestSetup;
+    VoidTest CustomOperatorFusion;
+    VoidTest CustomKernelWithBuiltInSchema;
+    VoidTest CustomKernelWithCustomSchema;
+};
+const CustomOpsTestApi& getapi();
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(CustomOpsScenarioTest, CustomOpsScenarioTestSetup)
+WINML_TEST(CustomOpsScenarioTest, CustomKernelWithBuiltInSchema)
+WINML_TEST(CustomOpsScenarioTest, CustomKernelWithCustomSchema)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(CustomOpsScenarioGpuTest, CustomOpsScenarioGpuTestSetup)
+WINML_TEST(CustomOpsScenarioGpuTest, CustomOperatorFusion)
+WINML_TEST_CLASS_END()
\ No newline at end of file
diff --git a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
index 5100dd08b71d1..1420b0c05e425 100644
--- a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
+++ b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
@@ -1,7 +1,6 @@
 #include "testPch.h"
 
 #include <d3dx12.h>
-#include <gtest/gtest.h>
 
 #include "winrt/Windows.Devices.Enumeration.Pnp.h"
 #include "winrt/Windows.Graphics.DirectX.Direct3D11.h"
@@ -20,7 +19,6 @@
 #include "DeviceHelpers.h"
 #include "filehelpers.h"
 #include "robuffer.h"
-#include "runtimeParameters.h"
 #include "Windows.AI.MachineLearning.Native.h"
 #include "Windows.Graphics.DirectX.Direct3D11.interop.h"
 #include "windows.ui.xaml.media.dxinterop.h"
@@ -30,7 +28,8 @@
 #include <d3d11.h>
 #include <initguid.h>
 #include <MemoryBuffer.h>
-
+#include "scenariotestscppwinrt.h"
+#include <iostream>
 #if __has_include("dxcore.h")
 #define ENABLE_DXCORE 1
 #endif
@@ -49,1408 +48,1285 @@ using namespace winrt::Windows::Storage;
 using namespace winrt::Windows::Storage::Streams;
 using namespace winrt::Windows::UI::Xaml::Media::Imaging;
 
-class ScenarioCppWinrtTest : public ::testing::Test
-{
-protected:
-    ScenarioCppWinrtTest()
-    {
-        init_apartment();
-    }
-};
+static void ScenarioCppWinrtTestSetup() {
+  init_apartment();
+}
 
-class ScenarioCppWinrtGpuTest : public ScenarioCppWinrtTest
-{
-protected:
-    void SetUp() override
-    {
-        GPUTEST
-    }
+static void ScenarioCppWinrtGpuTestSetup() {
+  ScenarioCppWinrtTestSetup();
+  GPUTEST
 };
-using ScenarioCppWinrtGpuTestDeathTest = ScenarioCppWinrtGpuTest;
-
-class ScenarioCppWinrtGpuSkipEdgeCoreTest : public ScenarioCppWinrtGpuTest
-{
-protected:
-    void SetUp() override
-    {
-        ScenarioCppWinrtGpuTest::SetUp();
-        SKIP_EDGECORE
-    }
+
+static void ScenarioCppWinrtGpuSkipEdgeCoreTestSetup() {
+  ScenarioCppWinrtGpuTestSetup();
+  SKIP_EDGECORE
 };
 
-TEST_F(ScenarioCppWinrtTest, Sample1)
-{
-    LearningModel model = nullptr;
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(filePath));
+static void Sample1() {
+  LearningModel model = nullptr;
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(filePath));
 }
 
-ILearningModelFeatureValue MakeTensor(const ITensorFeatureDescriptor& descriptor)
-{
-    auto dataType = descriptor.TensorKind();
-    std::vector<int64_t> shape;
-    int64_t size = 1;
-    for (auto&& dim : descriptor.Shape())
-    {
-        if (dim == -1) dim = 1;
-        shape.push_back(dim);
-        size *= dim;
-    }
+ILearningModelFeatureValue MakeTensor(const ITensorFeatureDescriptor& descriptor) {
+  auto dataType = descriptor.TensorKind();
+  std::vector<int64_t> shape;
+  int64_t size = 1;
+  for (auto&& dim : descriptor.Shape()) {
+    if (dim == -1) dim = 1;
+    shape.push_back(dim);
+    size *= dim;
+  }
 
-    switch (dataType)
-    {
-    case TensorKind::Float:
-    {
-        std::vector<float> buffer;
-        buffer.resize(size);
-        auto ftv = TensorFloat::CreateFromIterable(shape, winrt::single_threaded_vector<float>(std::move(buffer)));
-        return ftv;
+  switch (dataType) {
+    case TensorKind::Float: {
+      std::vector<float> buffer;
+      buffer.resize(size);
+      auto ftv = TensorFloat::CreateFromIterable(shape, winrt::single_threaded_vector<float>(std::move(buffer)));
+      return ftv;
     }
     default:
-        throw_hresult(E_NOTIMPL);
-        break;
-    }
+      throw_hresult(E_NOTIMPL);
+      break;
+  }
 }
 
-ILearningModelFeatureValue MakeImage(const IImageFeatureDescriptor& /*descriptor*/, winrt::Windows::Foundation::IInspectable data)
-{
-    VideoFrame videoFrame = nullptr;
-    if (data != nullptr)
-    {
-        SoftwareBitmap sb = nullptr;
-        data.as(sb);
-        videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
-    }
-    else
-    {
-        SoftwareBitmap sb = SoftwareBitmap(BitmapPixelFormat::Bgra8, 28, 28);
-        videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
-    }
-    auto imageValue = ImageFeatureValue::CreateFromVideoFrame(videoFrame);
-    return imageValue;
+ILearningModelFeatureValue MakeImage(const IImageFeatureDescriptor& /*descriptor*/, winrt::Windows::Foundation::IInspectable data) {
+  VideoFrame videoFrame = nullptr;
+  if (data != nullptr) {
+    SoftwareBitmap sb = nullptr;
+    data.as(sb);
+    videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
+  } else {
+    SoftwareBitmap sb = SoftwareBitmap(BitmapPixelFormat::Bgra8, 28, 28);
+    videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
+  }
+  auto imageValue = ImageFeatureValue::CreateFromVideoFrame(videoFrame);
+  return imageValue;
 }
 
-ILearningModelFeatureValue FeatureValueFromFeatureValueDescriptor(ILearningModelFeatureDescriptor descriptor, winrt::Windows::Foundation::IInspectable data = nullptr)
-{
-    auto kind = descriptor.Kind();
-    switch (kind)
-    {
-    case LearningModelFeatureKind::Image:
-    {
-        ImageFeatureDescriptor imageDescriptor = nullptr;
-        descriptor.as(imageDescriptor);
-        return MakeImage(imageDescriptor, data);
+ILearningModelFeatureValue FeatureValueFromFeatureValueDescriptor(ILearningModelFeatureDescriptor descriptor, winrt::Windows::Foundation::IInspectable data = nullptr) {
+  auto kind = descriptor.Kind();
+  switch (kind) {
+    case LearningModelFeatureKind::Image: {
+      ImageFeatureDescriptor imageDescriptor = nullptr;
+      descriptor.as(imageDescriptor);
+      return MakeImage(imageDescriptor, data);
     }
     case LearningModelFeatureKind::Map:
-        throw_hresult(E_NOTIMPL);
-        break;
+      throw_hresult(E_NOTIMPL);
+      break;
     case LearningModelFeatureKind::Sequence:
-        throw_hresult(E_NOTIMPL);
-        break;
-    case LearningModelFeatureKind::Tensor:
-    {
-        TensorFeatureDescriptor tensorDescriptor = nullptr;
-        descriptor.as(tensorDescriptor);
-        return MakeTensor(tensorDescriptor);
+      throw_hresult(E_NOTIMPL);
+      break;
+    case LearningModelFeatureKind::Tensor: {
+      TensorFeatureDescriptor tensorDescriptor = nullptr;
+      descriptor.as(tensorDescriptor);
+      return MakeTensor(tensorDescriptor);
     }
     default:
-        throw_hresult(E_INVALIDARG);
-        break;
-    }
+      throw_hresult(E_INVALIDARG);
+      break;
+  }
 }
 
 // helper method that populates a binding object with default data
-static void BindFeatures(LearningModelBinding binding, IVectorView<ILearningModelFeatureDescriptor> features)
-{
-    for (auto&& feature : features)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(feature);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        binding.Bind(feature.Name(), featureValue);
-    }
+static void BindFeatures(LearningModelBinding binding, IVectorView<ILearningModelFeatureDescriptor> features) {
+  for (auto&& feature : features) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(feature);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    binding.Bind(feature.Name(), featureValue);
+  }
 }
 
 //! Scenario1 : Load , bind, eval a model using all the system defaults (easy path)
-TEST_F(ScenarioCppWinrtTest, Scenario1LoadBindEvalDefault)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        binding.Bind(input.Name(), featureValue);
-    }
-    // run eval
-    EXPECT_NO_THROW(session.Evaluate(binding, L""));
+static void Scenario1LoadBindEvalDefault() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    binding.Bind(input.Name(), featureValue);
+  }
+  // run eval
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
 //! Scenario2: Load a model from stream
 //          - winRT, and win32
-TEST_F(ScenarioCppWinrtTest, Scenario2LoadModelFromStream)
-{
-    // get a stream
-    std::wstring path = FileHelpers::GetModulePath() + L"model.onnx";
-    auto storageFile = StorageFile::GetFileFromPathAsync(path).get();
-
-    // load the stream
-    Streams::IRandomAccessStreamReference streamref;
-    storageFile.as(streamref);
-
-    // load a model
-    LearningModel model = nullptr;
-    EXPECT_NO_THROW(model = LearningModel::LoadFromStreamAsync(streamref).get());
-    EXPECT_TRUE(model != nullptr);
+static void Scenario2LoadModelFromStream() {
+  // get a stream
+  std::wstring path = FileHelpers::GetModulePath() + L"model.onnx";
+  auto storageFile = StorageFile::GetFileFromPathAsync(path).get();
+
+  // load the stream
+  Streams::IRandomAccessStreamReference streamref;
+  storageFile.as(streamref);
+
+  // load a model
+  LearningModel model = nullptr;
+  WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromStreamAsync(streamref).get());
+  WINML_EXPECT_TRUE(model != nullptr);
 }
 
 //! Scenario3: pass a SoftwareBitmap into a model
-TEST_F(ScenarioCppWinrtGpuTest, Scenario3SoftwareBitmapInputBinding)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        // load the SoftwareBitmap
-        SoftwareBitmap sb = FileHelpers::GetSoftwareBitmapFromFile(FileHelpers::GetModulePath() + L"fish.png");
-        auto videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
-        auto imageValue = ImageFeatureValue::CreateFromVideoFrame(videoFrame);
-
-        EXPECT_NO_THROW(binding.Bind(input.Name(), imageValue));
-    }
-    // run eval
-    EXPECT_NO_THROW(session.Evaluate(binding, L""));
+static void Scenario3SoftwareBitmapInputBinding() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    // load the SoftwareBitmap
+    SoftwareBitmap sb = FileHelpers::GetSoftwareBitmapFromFile(FileHelpers::GetModulePath() + L"fish.png");
+    auto videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
+    auto imageValue = ImageFeatureValue::CreateFromVideoFrame(videoFrame);
+
+    WINML_EXPECT_NO_THROW(binding.Bind(input.Name(), imageValue));
+  }
+  // run eval
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
 //! Scenario5: run an async eval
-winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult> DoEvalAsync()
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        binding.Bind(input.Name(), featureValue);
-    }
-    // run eval async
-    return session.EvaluateAsync(binding, L"");
+winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult> DoEvalAsync() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    binding.Bind(input.Name(), featureValue);
+  }
+  // run eval async
+  return session.EvaluateAsync(binding, L"");
 }
 
-TEST_F(ScenarioCppWinrtTest, Scenario5AsyncEval)
-{
-    auto task = DoEvalAsync();
+static void Scenario5AsyncEval() {
+  auto task = DoEvalAsync();
 
-    while (task.Status() == winrt::Windows::Foundation::AsyncStatus::Started)
-    {
-        std::cout << "Waiting...\n";
-        Sleep(30);
-    }
-    std::cout << "Done\n";
-    EXPECT_NO_THROW(task.get());
+  while (task.Status() == winrt::Windows::Foundation::AsyncStatus::Started) {
+    std::cout << "Waiting...\n";
+    Sleep(30);
+  }
+  std::cout << "Done\n";
+  WINML_EXPECT_NO_THROW(task.get());
 }
 
 //! Scenario6: use BindInputWithProperties - BitmapBounds, BitmapPixelFormat
 // apparently this scenario is cut for rs5. - not cut, just rewprked. move props
 // to the image value when that is checked in.
-TEST_F(ScenarioCppWinrtGpuTest, Scenario6BindWithProperties)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        SoftwareBitmap sb = SoftwareBitmap(BitmapPixelFormat::Bgra8, 224, 224);
-        auto videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
-        auto imageValue = ImageFeatureValue::CreateFromVideoFrame(videoFrame);
-
-        PropertySet propertySet;
-
-        // make a BitmapBounds
-        BitmapBounds bounds;
-        bounds.X = 0;
-        bounds.Y = 0;
-        bounds.Height = 100;
-        bounds.Width = 100;
-
-        auto bitmapsBoundsProperty = winrt::Windows::Foundation::PropertyValue::CreateUInt32Array({ bounds.X, bounds.Y, bounds.Width, bounds.Height });
-        // insert it in the property set
-        propertySet.Insert(L"BitmapBounds", bitmapsBoundsProperty);
-
-        // make a BitmapPixelFormat
-        BitmapPixelFormat bitmapPixelFormat = BitmapPixelFormat::Bgra8;
-        // translate it to an int so it can be used as a PropertyValue;
-        int intFromBitmapPixelFormat = static_cast<int>(bitmapPixelFormat);
-        auto bitmapPixelFormatProperty = winrt::Windows::Foundation::PropertyValue::CreateInt32(intFromBitmapPixelFormat);
-        // insert it in the property set
-        propertySet.Insert(L"BitmapPixelFormat", bitmapPixelFormatProperty);
-
-        // bind with properties
-        EXPECT_NO_THROW(binding.Bind(input.Name(), imageValue, propertySet));
-    }
-    // run eval
-    EXPECT_NO_THROW(session.Evaluate(binding, L""));
+static void Scenario6BindWithProperties() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    SoftwareBitmap sb = SoftwareBitmap(BitmapPixelFormat::Bgra8, 224, 224);
+    auto videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
+    auto imageValue = ImageFeatureValue::CreateFromVideoFrame(videoFrame);
+
+    PropertySet propertySet;
+
+    // make a BitmapBounds
+    BitmapBounds bounds;
+    bounds.X = 0;
+    bounds.Y = 0;
+    bounds.Height = 100;
+    bounds.Width = 100;
+
+    auto bitmapsBoundsProperty = winrt::Windows::Foundation::PropertyValue::CreateUInt32Array({bounds.X, bounds.Y, bounds.Width, bounds.Height});
+    // insert it in the property set
+    propertySet.Insert(L"BitmapBounds", bitmapsBoundsProperty);
+
+    // make a BitmapPixelFormat
+    BitmapPixelFormat bitmapPixelFormat = BitmapPixelFormat::Bgra8;
+    // translate it to an int so it can be used as a PropertyValue;
+    int intFromBitmapPixelFormat = static_cast<int>(bitmapPixelFormat);
+    auto bitmapPixelFormatProperty = winrt::Windows::Foundation::PropertyValue::CreateInt32(intFromBitmapPixelFormat);
+    // insert it in the property set
+    propertySet.Insert(L"BitmapPixelFormat", bitmapPixelFormatProperty);
+
+    // bind with properties
+    WINML_EXPECT_NO_THROW(binding.Bind(input.Name(), imageValue, propertySet));
+  }
+  // run eval
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
 //! Scenario7: run eval without creating a binding object
-TEST_F(ScenarioCppWinrtTest, Scenario7EvalWithNoBind)
-{
-    auto map = winrt::single_threaded_map<hstring, winrt::Windows::Foundation::IInspectable>();
-
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // enumerate feature descriptors and create features (but don't bind them)
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
-        map.Insert(input.Name(), featureValue);
-    }
-    // run eval
-    EXPECT_NO_THROW(session.EvaluateFeaturesAsync(map, L"").get());
+static void Scenario7EvalWithNoBind() {
+  auto map = winrt::single_threaded_map<hstring, winrt::Windows::Foundation::IInspectable>();
+
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // enumerate feature descriptors and create features (but don't bind them)
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
+    map.Insert(input.Name(), featureValue);
+  }
+  // run eval
+  WINML_EXPECT_NO_THROW(session.EvaluateFeaturesAsync(map, L"").get());
 }
 
 //! Scenario8: choose which device to run the model on - PreferredDeviceType, PreferredDevicePerformance, SetDeviceFromSurface, SetDevice
 // create a session on the default device
-TEST_F(ScenarioCppWinrtTest, Scenario8SetDeviceSampleDefault)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    LearningModelDevice anyDevice(LearningModelDeviceKind::Default);
-    LearningModelSession anySession(model, anyDevice);
+static void Scenario8SetDeviceSampleDefault() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  LearningModelDevice anyDevice(LearningModelDeviceKind::Default);
+  LearningModelSession anySession(model, anyDevice);
 }
 
 // create a session on the CPU device
-TEST_F(ScenarioCppWinrtTest, Scenario8SetDeviceSampleCPU)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    LearningModelDevice cpuDevice(LearningModelDeviceKind::Cpu);
-    LearningModelSession cpuSession(model, cpuDevice);
+static void Scenario8SetDeviceSampleCPU() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  LearningModelDevice cpuDevice(LearningModelDeviceKind::Cpu);
+  LearningModelSession cpuSession(model, cpuDevice);
 }
 
 // create a session on the default DML device
-TEST_F(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleDefaultDirectX)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    LearningModelDevice dmlDeviceDefault(LearningModelDeviceKind::DirectX);
-    LearningModelSession dmlSessionDefault(model, dmlDeviceDefault);
+static void Scenario8SetDeviceSampleDefaultDirectX() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  LearningModelDevice dmlDeviceDefault(LearningModelDeviceKind::DirectX);
+  LearningModelSession dmlSessionDefault(model, dmlDeviceDefault);
 }
 
 // create a session on the DML device that provides best power
-TEST_F(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMinPower)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    LearningModelDevice dmlDeviceMinPower(LearningModelDeviceKind::DirectXMinPower);
-    LearningModelSession dmlSessionMinPower(model, dmlDeviceMinPower);
+static void Scenario8SetDeviceSampleMinPower() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  LearningModelDevice dmlDeviceMinPower(LearningModelDeviceKind::DirectXMinPower);
+  LearningModelSession dmlSessionMinPower(model, dmlDeviceMinPower);
 }
 
 // create a session on the DML device that provides best perf
-TEST_F(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMaxPerf)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    LearningModelDevice dmlDeviceMaxPerf(LearningModelDeviceKind::DirectXHighPerformance);
-    LearningModelSession dmlSessionMaxPerf(model, dmlDeviceMaxPerf);
+static void Scenario8SetDeviceSampleMaxPerf() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  LearningModelDevice dmlDeviceMaxPerf(LearningModelDeviceKind::DirectXHighPerformance);
+  LearningModelSession dmlSessionMaxPerf(model, dmlDeviceMaxPerf);
 }
 
 // create a session on the same device my camera is on
-TEST_F(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMyCameraDevice)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    auto devices = winrt::Windows::Devices::Enumeration::DeviceInformation::FindAllAsync(winrt::Windows::Devices::Enumeration::DeviceClass::VideoCapture).get();
-    hstring deviceId;
-    if (devices.Size() > 0)
-    {
-        auto device = devices.GetAt(0);
-        deviceId = device.Id();
-        auto deviceName = device.Name();
-        auto enabled = device.IsEnabled();
-        std::cout << "Found device " << deviceName.c_str() << ", enabled = " << enabled << "\n";
-        winrt::Windows::Media::Capture::MediaCapture captureManager;
-        winrt::Windows::Media::Capture::MediaCaptureInitializationSettings settings;
-        settings.VideoDeviceId(deviceId);
-        captureManager.InitializeAsync(settings).get();
-        auto mediaCaptureSettings = captureManager.MediaCaptureSettings();
-        auto direct3D11Device = mediaCaptureSettings.Direct3D11Device();
-        LearningModelDevice dmlDeviceCamera = LearningModelDevice::CreateFromDirect3D11Device(direct3D11Device);
-        LearningModelSession dmlSessionCamera(model, dmlDeviceCamera);
-    }
-    else
-    {
-        GTEST_SKIP() << "Test skipped because video capture device is missing";
-    }
+static void Scenario8SetDeviceSampleMyCameraDevice() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  auto devices = winrt::Windows::Devices::Enumeration::DeviceInformation::FindAllAsync(winrt::Windows::Devices::Enumeration::DeviceClass::VideoCapture).get();
+  hstring deviceId;
+  if (devices.Size() > 0) {
+    auto device = devices.GetAt(0);
+    deviceId = device.Id();
+    auto deviceName = device.Name();
+    auto enabled = device.IsEnabled();
+    std::cout << "Found device " << deviceName.c_str() << ", enabled = " << enabled << "\n";
+    winrt::Windows::Media::Capture::MediaCapture captureManager;
+    winrt::Windows::Media::Capture::MediaCaptureInitializationSettings settings;
+    settings.VideoDeviceId(deviceId);
+    captureManager.InitializeAsync(settings).get();
+    auto mediaCaptureSettings = captureManager.MediaCaptureSettings();
+    auto direct3D11Device = mediaCaptureSettings.Direct3D11Device();
+    LearningModelDevice dmlDeviceCamera = LearningModelDevice::CreateFromDirect3D11Device(direct3D11Device);
+    LearningModelSession dmlSessionCamera(model, dmlDeviceCamera);
+  } else {
+    WINML_SKIP_TEST("Test skipped because video capture device is missing");
+  }
 }
 
 // create a device from D3D11 Device
-TEST_F(ScenarioCppWinrtGpuSkipEdgeCoreTest, Scenario8SetDeviceSampleD3D11Device)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    com_ptr<ID3D11Device> pD3D11Device = nullptr;
-    com_ptr<ID3D11DeviceContext> pContext = nullptr;
-    D3D_FEATURE_LEVEL fl;
-    HRESULT result = D3D11CreateDevice(
-        nullptr, D3D_DRIVER_TYPE::D3D_DRIVER_TYPE_HARDWARE, nullptr, 0, nullptr, 0,
-        D3D11_SDK_VERSION, pD3D11Device.put(), &fl, pContext.put());
-    if (FAILED(result))
-    {
-        GTEST_SKIP() << "Test skipped because d3d11 device is missing";
-    }
+static void Scenario8SetDeviceSampleD3D11Device() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  com_ptr<ID3D11Device> pD3D11Device = nullptr;
+  com_ptr<ID3D11DeviceContext> pContext = nullptr;
+  D3D_FEATURE_LEVEL fl;
+  HRESULT result = D3D11CreateDevice(
+      nullptr, D3D_DRIVER_TYPE::D3D_DRIVER_TYPE_HARDWARE, nullptr, 0, nullptr, 0,
+      D3D11_SDK_VERSION, pD3D11Device.put(), &fl, pContext.put());
+  if (FAILED(result)) {
+    WINML_SKIP_TEST("Test skipped because d3d11 device is missing");
+  }
 
-    // get dxgiDevice from d3ddevice
-    com_ptr<IDXGIDevice> pDxgiDevice;
-    pD3D11Device.get()->QueryInterface<IDXGIDevice>(pDxgiDevice.put());
+  // get dxgiDevice from d3ddevice
+  com_ptr<IDXGIDevice> pDxgiDevice;
+  pD3D11Device.get()->QueryInterface<IDXGIDevice>(pDxgiDevice.put());
 
-    com_ptr<::IInspectable> pInspectable;
-    CreateDirect3D11DeviceFromDXGIDevice(pDxgiDevice.get(), pInspectable.put());
+  com_ptr<::IInspectable> pInspectable;
+  CreateDirect3D11DeviceFromDXGIDevice(pDxgiDevice.get(), pInspectable.put());
 
-    LearningModelDevice device = LearningModelDevice::CreateFromDirect3D11Device(
-        pInspectable.as<winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice>());
-    LearningModelSession session(model, device);
+  LearningModelDevice device = LearningModelDevice::CreateFromDirect3D11Device(
+      pInspectable.as<winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice>());
+  LearningModelSession session(model, device);
 }
 
 // create a session on the a specific dx device that I chose some other way , note we have to use native interop here and pass a cmd queue
-TEST_F(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleCustomCommandQueue)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    com_ptr<ID3D12Device> pD3D12Device = nullptr;
-    DeviceHelpers::AdapterEnumerationSupport support;
-    if (FAILED(DeviceHelpers::GetAdapterEnumerationSupport(&support)))
-    {
-        FAIL() << "Unable to load DXGI or DXCore";
-        return;
-    }
-    HRESULT result = S_OK;
-    if (support.has_dxgi)
-    {
-        EXPECT_NO_THROW(result = D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_12_0, __uuidof(ID3D12Device), reinterpret_cast<void**>(pD3D12Device.put())));
-    }
+static void Scenario8SetDeviceSampleCustomCommandQueue() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+
+  com_ptr<ID3D12Device> pD3D12Device = nullptr;
+  DeviceHelpers::AdapterEnumerationSupport support;
+  if (FAILED(DeviceHelpers::GetAdapterEnumerationSupport(&support))) {
+    WINML_LOG_ERROR("Unable to load DXGI or DXCore");
+    return;
+  }
+  HRESULT result = S_OK;
+  if (support.has_dxgi) {
+    WINML_EXPECT_NO_THROW(result = D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_12_0, __uuidof(ID3D12Device), reinterpret_cast<void**>(pD3D12Device.put())));
+  }
 #ifdef ENABLE_DXCORE
-    if (support.has_dxgi == false)
-    {
-        com_ptr<IDXCoreAdapterFactory> spFactory;
-        DXCoreCreateAdapterFactory(IID_PPV_ARGS(spFactory.put()));
-        const GUID gpuFilter[] = { DXCORE_ADAPTER_ATTRIBUTE_D3D12_GRAPHICS };
-        com_ptr<IDXCoreAdapterList> spAdapterList;
-        spFactory->CreateAdapterList(1, gpuFilter, IID_PPV_ARGS(spAdapterList.put()));
-        com_ptr<IDXCoreAdapter> spAdapter;
-        EXPECT_NO_THROW(spAdapterList->GetAdapter(0, IID_PPV_ARGS(spAdapter.put())));
-        ::IUnknown* pAdapter = spAdapter.get();
-        EXPECT_NO_THROW(result = D3D12CreateDevice(pAdapter, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_12_0, __uuidof(ID3D12Device), reinterpret_cast<void**>(pD3D12Device.put())));
-    }
+  if (support.has_dxgi == false) {
+    com_ptr<IDXCoreAdapterFactory> spFactory;
+    DXCoreCreateAdapterFactory(IID_PPV_ARGS(spFactory.put()));
+    const GUID gpuFilter[] = {DXCORE_ADAPTER_ATTRIBUTE_D3D12_GRAPHICS};
+    com_ptr<IDXCoreAdapterList> spAdapterList;
+    spFactory->CreateAdapterList(1, gpuFilter, IID_PPV_ARGS(spAdapterList.put()));
+    com_ptr<IDXCoreAdapter> spAdapter;
+    WINML_EXPECT_NO_THROW(spAdapterList->GetAdapter(0, IID_PPV_ARGS(spAdapter.put())));
+    ::IUnknown* pAdapter = spAdapter.get();
+    WINML_EXPECT_NO_THROW(result = D3D12CreateDevice(pAdapter, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_12_0, __uuidof(ID3D12Device), reinterpret_cast<void**>(pD3D12Device.put())));
+  }
 #endif
 
-    if (FAILED(result))
-    {
-        GTEST_SKIP() << "Test skipped because d3d12 device is missing";
-        return;
-    }
-    com_ptr<ID3D12CommandQueue> dxQueue = nullptr;
-    D3D12_COMMAND_QUEUE_DESC commandQueueDesc = {};
-    commandQueueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
-    pD3D12Device->CreateCommandQueue(&commandQueueDesc, __uuidof(ID3D12CommandQueue), reinterpret_cast<void**>(&dxQueue));
-    auto factory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
+  if (FAILED(result)) {
+    WINML_SKIP_TEST("Test skipped because d3d12 device is missing");
+    return;
+  }
+  com_ptr<ID3D12CommandQueue> dxQueue = nullptr;
+  D3D12_COMMAND_QUEUE_DESC commandQueueDesc = {};
+  commandQueueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
+  pD3D12Device->CreateCommandQueue(&commandQueueDesc, __uuidof(ID3D12CommandQueue), reinterpret_cast<void**>(&dxQueue));
+  auto factory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
 
-    com_ptr<::IUnknown> spUnk;
-    factory->CreateFromD3D12CommandQueue(dxQueue.get(), spUnk.put());
+  com_ptr<::IUnknown> spUnk;
+  factory->CreateFromD3D12CommandQueue(dxQueue.get(), spUnk.put());
 
-    auto dmlDeviceCustom = spUnk.as<LearningModelDevice>();
-    LearningModelSession dmlSessionCustom(model, dmlDeviceCustom);
+  auto dmlDeviceCustom = spUnk.as<LearningModelDevice>();
+  LearningModelSession dmlSessionCustom(model, dmlDeviceCustom);
 }
 
-
 //pass a Tensor in as an input GPU
-TEST_F(ScenarioCppWinrtGpuTest, DISABLED_Scenario9LoadBindEvalInputTensorGPU)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"fns-candy.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-
-    com_ptr<ID3D12Device> pD3D12Device;
-    EXPECT_NO_THROW(D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device), pD3D12Device.put_void()));
-    com_ptr<ID3D12CommandQueue> dxQueue;
-    D3D12_COMMAND_QUEUE_DESC commandQueueDesc = {};
-    commandQueueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
-    pD3D12Device->CreateCommandQueue(&commandQueueDesc, __uuidof(ID3D12CommandQueue), dxQueue.put_void());
-    auto devicefactory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
-    auto tensorfactory = get_activation_factory<TensorFloat, ITensorStaticsNative>();
-
-
-    com_ptr<::IUnknown> spUnk;
-    EXPECT_NO_THROW(devicefactory->CreateFromD3D12CommandQueue(dxQueue.get(), spUnk.put()));
-
-    LearningModelDevice dmlDeviceCustom = nullptr;
-    EXPECT_NO_THROW(spUnk.as(dmlDeviceCustom));
-    LearningModelSession dmlSessionCustom = nullptr;
-    EXPECT_NO_THROW(dmlSessionCustom = LearningModelSession(model, dmlDeviceCustom));
-
-    LearningModelBinding modelBinding(dmlSessionCustom);
-
-    UINT64 bufferbytesize = 720 * 720 * 3 * sizeof(float);
-    D3D12_HEAP_PROPERTIES heapProperties = {
-        D3D12_HEAP_TYPE_DEFAULT,
-        D3D12_CPU_PAGE_PROPERTY_UNKNOWN,
-        D3D12_MEMORY_POOL_UNKNOWN,
-        0,
-        0
-    };
-    D3D12_RESOURCE_DESC resourceDesc = {
-        D3D12_RESOURCE_DIMENSION_BUFFER,
-        0,
-        bufferbytesize,
-        1,
-        1,
-        1,
-        DXGI_FORMAT_UNKNOWN,
-    { 1, 0 },
-    D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
-    D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS
-    };
-
-    com_ptr<ID3D12Resource> pGPUResource = nullptr;
-    pD3D12Device->CreateCommittedResource(
-        &heapProperties,
-        D3D12_HEAP_FLAG_NONE,
-        &resourceDesc,
-        D3D12_RESOURCE_STATE_COMMON,
-        nullptr,
-        __uuidof(ID3D12Resource),
-        pGPUResource.put_void()
-    );
-    com_ptr<::IUnknown> spUnkTensor;
-    TensorFloat input1imagetensor(nullptr);
-    __int64 shape[4] = { 1,3, 720, 720 };
-    tensorfactory->CreateFromD3D12Resource(pGPUResource.get(), shape, 4, spUnkTensor.put());
-    spUnkTensor.try_as(input1imagetensor);
-
-    auto feature = model.InputFeatures().First();
-    EXPECT_NO_THROW(modelBinding.Bind(feature.Current().Name(), input1imagetensor));
-
-    auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
-    auto outputtensorshape = outputtensordescriptor.Shape();
-    VideoFrame outputimage(
-		BitmapPixelFormat::Rgba8,
-		static_cast<int32_t>(outputtensorshape.GetAt(3)),
-		static_cast<int32_t>(outputtensorshape.GetAt(2)));
-    ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
-
-    EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
-
-    // Testing GetAsD3D12Resource
-    com_ptr<ID3D12Resource> pReturnedResource;
-    input1imagetensor.as<ITensorNative>()->GetD3D12Resource(pReturnedResource.put());
-    EXPECT_EQ(pReturnedResource.get(), pGPUResource.get());
-
-    // Evaluate the model
-    winrt::hstring correlationId;
-    dmlSessionCustom.EvaluateAsync(modelBinding, correlationId).get();
+static void Scenario9LoadBindEvalInputTensorGPU() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"fns-candy.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
 
+  com_ptr<ID3D12Device> pD3D12Device;
+  WINML_EXPECT_NO_THROW(D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device), pD3D12Device.put_void()));
+  com_ptr<ID3D12CommandQueue> dxQueue;
+  D3D12_COMMAND_QUEUE_DESC commandQueueDesc = {};
+  commandQueueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
+  pD3D12Device->CreateCommandQueue(&commandQueueDesc, __uuidof(ID3D12CommandQueue), dxQueue.put_void());
+  auto devicefactory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
+  auto tensorfactory = get_activation_factory<TensorFloat, ITensorStaticsNative>();
+
+  com_ptr<::IUnknown> spUnk;
+  WINML_EXPECT_NO_THROW(devicefactory->CreateFromD3D12CommandQueue(dxQueue.get(), spUnk.put()));
+
+  LearningModelDevice dmlDeviceCustom = nullptr;
+  WINML_EXPECT_NO_THROW(spUnk.as(dmlDeviceCustom));
+  LearningModelSession dmlSessionCustom = nullptr;
+  WINML_EXPECT_NO_THROW(dmlSessionCustom = LearningModelSession(model, dmlDeviceCustom));
+
+  LearningModelBinding modelBinding(dmlSessionCustom);
+
+  UINT64 bufferbytesize = 720 * 720 * 3 * sizeof(float);
+  D3D12_HEAP_PROPERTIES heapProperties = {
+      D3D12_HEAP_TYPE_DEFAULT,
+      D3D12_CPU_PAGE_PROPERTY_UNKNOWN,
+      D3D12_MEMORY_POOL_UNKNOWN,
+      0,
+      0};
+  D3D12_RESOURCE_DESC resourceDesc = {
+      D3D12_RESOURCE_DIMENSION_BUFFER,
+      0,
+      bufferbytesize,
+      1,
+      1,
+      1,
+      DXGI_FORMAT_UNKNOWN,
+      {1, 0},
+      D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
+      D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS};
+
+  com_ptr<ID3D12Resource> pGPUResource = nullptr;
+  pD3D12Device->CreateCommittedResource(
+      &heapProperties,
+      D3D12_HEAP_FLAG_NONE,
+      &resourceDesc,
+      D3D12_RESOURCE_STATE_COMMON,
+      nullptr,
+      __uuidof(ID3D12Resource),
+      pGPUResource.put_void());
+  com_ptr<::IUnknown> spUnkTensor;
+  TensorFloat input1imagetensor(nullptr);
+  __int64 shape[4] = {1, 3, 720, 720};
+  tensorfactory->CreateFromD3D12Resource(pGPUResource.get(), shape, 4, spUnkTensor.put());
+  spUnkTensor.try_as(input1imagetensor);
+
+  auto feature = model.InputFeatures().First();
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(feature.Current().Name(), input1imagetensor));
+
+  auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
+  auto outputtensorshape = outputtensordescriptor.Shape();
+  VideoFrame outputimage(
+      BitmapPixelFormat::Rgba8,
+      static_cast<int32_t>(outputtensorshape.GetAt(3)),
+      static_cast<int32_t>(outputtensorshape.GetAt(2)));
+  ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
+
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
+
+  // Testing GetAsD3D12Resource
+  com_ptr<ID3D12Resource> pReturnedResource;
+  input1imagetensor.as<ITensorNative>()->GetD3D12Resource(pReturnedResource.put());
+  WINML_EXPECT_EQUAL(pReturnedResource.get(), pGPUResource.get());
+
+  // Evaluate the model
+  winrt::hstring correlationId;
+  dmlSessionCustom.EvaluateAsync(modelBinding, correlationId).get();
 }
 
-TEST_F(ScenarioCppWinrtGpuTest, Scenario13SingleModelOnCPUandGPU)
-{
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    LearningModelSession cpuSession(model, LearningModelDevice(LearningModelDeviceKind::Cpu));
-    LearningModelSession gpuSession(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
-
-    LearningModelBinding cpuBinding(cpuSession);
-    LearningModelBinding gpuBinding(gpuSession);
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto cpuFeatureValue = FeatureValueFromFeatureValueDescriptor(input);
-        cpuBinding.Bind(input.Name(), cpuFeatureValue);
-
-        auto gpuFeatureValue = FeatureValueFromFeatureValueDescriptor(input);
-        gpuBinding.Bind(input.Name(), gpuFeatureValue);
-    }
+static void Scenario13SingleModelOnCPUandGPU() {
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  LearningModelSession cpuSession(model, LearningModelDevice(LearningModelDeviceKind::Cpu));
+  LearningModelSession gpuSession(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
+
+  LearningModelBinding cpuBinding(cpuSession);
+  LearningModelBinding gpuBinding(gpuSession);
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto cpuFeatureValue = FeatureValueFromFeatureValueDescriptor(input);
+    cpuBinding.Bind(input.Name(), cpuFeatureValue);
+
+    auto gpuFeatureValue = FeatureValueFromFeatureValueDescriptor(input);
+    gpuBinding.Bind(input.Name(), gpuFeatureValue);
+  }
 
-    auto cpuTask = cpuSession.EvaluateAsync(cpuBinding, L"cpu");
-    auto gpuTask = gpuSession.EvaluateAsync(gpuBinding, L"gpu");
+  auto cpuTask = cpuSession.EvaluateAsync(cpuBinding, L"cpu");
+  auto gpuTask = gpuSession.EvaluateAsync(gpuBinding, L"gpu");
 
-    EXPECT_NO_THROW(cpuTask.get());
-    EXPECT_NO_THROW(gpuTask.get());
+  WINML_EXPECT_NO_THROW(cpuTask.get());
+  WINML_EXPECT_NO_THROW(gpuTask.get());
 }
 
 // Validates when binding input image with free dimensions, the binding step is executed correctly.
-TEST_F(ScenarioCppWinrtGpuTest, Scenario11FreeDimensionsTensor)
-{
-    std::wstring filePath = FileHelpers::GetModulePath() + L"free_dimensional_image_input.onnx";
-    // load a model with expected input size: -1 x -1
-    auto model = LearningModel::LoadFromFilePath(filePath);
-    auto session = LearningModelSession(model);
-    auto binding = LearningModelBinding(session);
+static void Scenario11FreeDimensionsTensor() {
+  std::wstring filePath = FileHelpers::GetModulePath() + L"free_dimensional_image_input.onnx";
+  // load a model with expected input size: -1 x -1
+  auto model = LearningModel::LoadFromFilePath(filePath);
+  auto session = LearningModelSession(model);
+  auto binding = LearningModelBinding(session);
 
-    VideoFrame inputImage(BitmapPixelFormat::Rgba8, 1000, 1000);
-    ImageFeatureValue inputimagetensor = ImageFeatureValue::CreateFromVideoFrame(inputImage);
+  VideoFrame inputImage(BitmapPixelFormat::Rgba8, 1000, 1000);
+  ImageFeatureValue inputimagetensor = ImageFeatureValue::CreateFromVideoFrame(inputImage);
 
-    auto feature = model.InputFeatures().First();
-    binding.Bind(feature.Current().Name(), inputimagetensor);
-    feature.MoveNext();
-    binding.Bind(feature.Current().Name(), inputimagetensor);
+  auto feature = model.InputFeatures().First();
+  binding.Bind(feature.Current().Name(), inputimagetensor);
+  feature.MoveNext();
+  binding.Bind(feature.Current().Name(), inputimagetensor);
 
-    session.Evaluate(binding, L"");
+  session.Evaluate(binding, L"");
 }
 
-TEST_F(ScenarioCppWinrtGpuTest, Scenario11FreeDimensionsImage)
-{
-    std::wstring filePath = FileHelpers::GetModulePath() + L"free_dimensional_imageDes.onnx";
-    // load a model with expected input size: -1 x -1
-    auto model = LearningModel::LoadFromFilePath(filePath);
-    auto session = LearningModelSession(model);
-    auto binding = LearningModelBinding(session);
+static void Scenario11FreeDimensionsImage() {
+  std::wstring filePath = FileHelpers::GetModulePath() + L"free_dimensional_imageDes.onnx";
+  // load a model with expected input size: -1 x -1
+  auto model = LearningModel::LoadFromFilePath(filePath);
+  auto session = LearningModelSession(model);
+  auto binding = LearningModelBinding(session);
 
-    VideoFrame inputImage(BitmapPixelFormat::Bgra8, 1000, 1000);
-    ImageFeatureValue inputimagetensor = ImageFeatureValue::CreateFromVideoFrame(inputImage);
+  VideoFrame inputImage(BitmapPixelFormat::Bgra8, 1000, 1000);
+  ImageFeatureValue inputimagetensor = ImageFeatureValue::CreateFromVideoFrame(inputImage);
 
-    auto feature = model.InputFeatures().First();
-    ImageFeatureDescriptor imageDescriptor = nullptr;
-    feature.Current().as(imageDescriptor);
-    binding.Bind(feature.Current().Name(), inputimagetensor);
+  auto feature = model.InputFeatures().First();
+  ImageFeatureDescriptor imageDescriptor = nullptr;
+  feature.Current().as(imageDescriptor);
+  binding.Bind(feature.Current().Name(), inputimagetensor);
 
-    feature.MoveNext();
-    feature.Current().as(imageDescriptor);
-    binding.Bind(feature.Current().Name(), inputimagetensor);
+  feature.MoveNext();
+  feature.Current().as(imageDescriptor);
+  binding.Bind(feature.Current().Name(), inputimagetensor);
 
-    session.Evaluate(binding, L"");
+  session.Evaluate(binding, L"");
 }
 
-struct SwapChainEntry
-{
-    LearningModelSession session;
-    LearningModelBinding binding;
-    winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult> activetask;
-    SwapChainEntry() :session(nullptr), binding(nullptr), activetask(nullptr)
-    {}
+struct SwapChainEntry {
+  LearningModelSession session;
+  LearningModelBinding binding;
+  winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult> activetask;
+  SwapChainEntry() : session(nullptr), binding(nullptr), activetask(nullptr) {}
 };
-void SubmitEval(LearningModel model, SwapChainEntry *sessionBindings, int swapchaindex)
-{
-    if (sessionBindings[swapchaindex].activetask != nullptr)
-    {
-        //make sure the previously submitted work for this swapchain index is complete before reusing resources
-        sessionBindings[swapchaindex].activetask.get();
-    }
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        sessionBindings[swapchaindex].binding.Bind(input.Name(), featureValue);
-    }
-    // submit an eval and wait for it to finish submitting work
-    sessionBindings[swapchaindex].activetask = sessionBindings[swapchaindex].session.EvaluateAsync(sessionBindings[swapchaindex].binding, L"0");
-    // return without waiting for the submit to finish, setup the completion handler
+void SubmitEval(LearningModel model, SwapChainEntry* sessionBindings, int swapchaindex) {
+  if (sessionBindings[swapchaindex].activetask != nullptr) {
+    //make sure the previously submitted work for this swapchain index is complete before reusing resources
+    sessionBindings[swapchaindex].activetask.get();
+  }
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    sessionBindings[swapchaindex].binding.Bind(input.Name(), featureValue);
+  }
+  // submit an eval and wait for it to finish submitting work
+  sessionBindings[swapchaindex].activetask = sessionBindings[swapchaindex].session.EvaluateAsync(sessionBindings[swapchaindex].binding, L"0");
+  // return without waiting for the submit to finish, setup the completion handler
 }
 
 //Scenario14:Load single model, run it mutliple times on a single gpu device using a fast swapchain pattern
-TEST_F(ScenarioCppWinrtGpuTest, Scenario14RunModelSwapchain)
-{
-    const int swapchainentrycount = 3;
-    SwapChainEntry sessionBindings[swapchainentrycount];
-
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on gpu1
-    LearningModelDevice dmlDevice = LearningModelDevice(LearningModelDeviceKind::DirectX);
-    // create the swapchain style bindings to cycle through
-    for (int i = 0; i < swapchainentrycount; i++)
-    {
-        sessionBindings[i].session = LearningModelSession(model, dmlDevice);
-        sessionBindings[i].binding = LearningModelBinding(sessionBindings[i].session);
-    }
+static void Scenario14RunModelSwapchain() {
+  const int swapchainentrycount = 3;
+  SwapChainEntry sessionBindings[swapchainentrycount];
 
-    //submit 10 evaluations to 3 swapchain entries
-    int swapchaindex = 0;
-    for (int i = 0; i < 10; i++)
-    {
-        swapchaindex = swapchaindex % swapchainentrycount;
-        SubmitEval(model, sessionBindings, (swapchaindex)++);
-    }
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on gpu1
+  LearningModelDevice dmlDevice = LearningModelDevice(LearningModelDeviceKind::DirectX);
+  // create the swapchain style bindings to cycle through
+  for (int i = 0; i < swapchainentrycount; i++) {
+    sessionBindings[i].session = LearningModelSession(model, dmlDevice);
+    sessionBindings[i].binding = LearningModelBinding(sessionBindings[i].session);
+  }
+
+  //submit 10 evaluations to 3 swapchain entries
+  int swapchaindex = 0;
+  for (int i = 0; i < 10; i++) {
+    swapchaindex = swapchaindex % swapchainentrycount;
+    SubmitEval(model, sessionBindings, (swapchaindex)++);
+  }
 
-    //wait for all work to be completed
-    for (int i = 0; i < swapchainentrycount; i++)
-    {
-        if (sessionBindings[i].activetask != nullptr)
-        {
-            //make sure the previously submitted work for this swapchain index is compolete before resuing resources
-            sessionBindings[i].activetask.get();
-        }
+  //wait for all work to be completed
+  for (int i = 0; i < swapchainentrycount; i++) {
+    if (sessionBindings[i].activetask != nullptr) {
+      //make sure the previously submitted work for this swapchain index is compolete before resuing resources
+      sessionBindings[i].activetask.get();
     }
+  }
 }
-static void LoadBindEval_CustomOperator_CPU(const wchar_t* fileName)
-{
-    auto customOperatorProvider = winrt::make<CustomOperatorProvider>();
-    auto provider = customOperatorProvider.as<ILearningModelOperatorProvider>();
-
-    LearningModel model = LearningModel::LoadFromFilePath(fileName, provider);
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    LearningModelBinding bindings(session);
-
-    auto inputShape = std::vector<int64_t>{ 5 };
-    auto inputData = std::vector<float>{ -50.f, -25.f, 0.f, 25.f, 50.f };
-    auto inputValue =
-        TensorFloat::CreateFromIterable(
-            inputShape,
-            single_threaded_vector<float>(std::move(inputData)).GetView());
-    EXPECT_NO_THROW(bindings.Bind(L"X", inputValue));
-
-    auto outputValue = TensorFloat::Create();
-    EXPECT_NO_THROW(bindings.Bind(L"Y", outputValue));
-
-    hstring correlationId;
-    EXPECT_NO_THROW(session.Evaluate(bindings, correlationId));
-
-    auto buffer = outputValue.GetAsVectorView();
-    EXPECT_TRUE(buffer != nullptr);
+static void LoadBindEval_CustomOperator_CPU(const wchar_t* fileName) {
+  auto customOperatorProvider = winrt::make<CustomOperatorProvider>();
+  auto provider = customOperatorProvider.as<ILearningModelOperatorProvider>();
+
+  LearningModel model = LearningModel::LoadFromFilePath(fileName, provider);
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  LearningModelBinding bindings(session);
+
+  auto inputShape = std::vector<int64_t>{5};
+  auto inputData = std::vector<float>{-50.f, -25.f, 0.f, 25.f, 50.f};
+  auto inputValue =
+      TensorFloat::CreateFromIterable(
+          inputShape,
+          single_threaded_vector<float>(std::move(inputData)).GetView());
+  WINML_EXPECT_NO_THROW(bindings.Bind(L"X", inputValue));
+
+  auto outputValue = TensorFloat::Create();
+  WINML_EXPECT_NO_THROW(bindings.Bind(L"Y", outputValue));
+
+  hstring correlationId;
+  WINML_EXPECT_NO_THROW(session.Evaluate(bindings, correlationId));
+
+  auto buffer = outputValue.GetAsVectorView();
+  WINML_EXPECT_TRUE(buffer != nullptr);
 }
 
 //! Scenario17 : Control the dev diagnostics features of WinML Tracing
-TEST_F(ScenarioCppWinrtTest, Scenario17DevDiagnostics)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        binding.Bind(input.Name(), featureValue);
-    }
-    session.EvaluationProperties().Insert(L"EnableDebugOutput", nullptr);
-    // run eval
-    EXPECT_NO_THROW(session.Evaluate(binding, L""));
+static void Scenario17DevDiagnostics() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    binding.Bind(input.Name(), featureValue);
+  }
+  session.EvaluationProperties().Insert(L"EnableDebugOutput", nullptr);
+  // run eval
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
 /** 
  * Custom Operator Tests are labeled as GPU tests because the DML code is interlaced with the custom op code
  * even though CPU custom ops shouldn't be dependent on GPU functionality.
- * These should be reclassed to ScenarioCppWinrtTest once the DML code is decoupled from the custom op code.
-**/ 
-
+ * These should be reclassed to ScenarioCppWinrt once the DML code is decoupled from the custom op code.
+**/
 // create a session that loads a model with a branch new operator, register the custom operator, and load/bind/eval
-TEST_F(ScenarioCppWinrtGpuTest, Scenario20aLoadBindEvalCustomOperatorCPU) 
-{
-    std::wstring filePath = FileHelpers::GetModulePath() + L"noisy_relu.onnx";
-    LoadBindEval_CustomOperator_CPU(filePath.c_str());
+static void Scenario20aLoadBindEvalCustomOperatorCPU() {
+  std::wstring filePath = FileHelpers::GetModulePath() + L"noisy_relu.onnx";
+  LoadBindEval_CustomOperator_CPU(filePath.c_str());
 }
 
 // create a session that loads a model with an overridden operator, register the replacement custom operator, and load/bind/eval
-TEST_F(ScenarioCppWinrtGpuTest, Scenario20bLoadBindEvalReplacementCustomOperatorCPU)
-{
-    std::wstring filePath = FileHelpers::GetModulePath() + L"relu.onnx";
-    LoadBindEval_CustomOperator_CPU(filePath.c_str());
+static void Scenario20bLoadBindEvalReplacementCustomOperatorCPU() {
+  std::wstring filePath = FileHelpers::GetModulePath() + L"relu.onnx";
+  LoadBindEval_CustomOperator_CPU(filePath.c_str());
 }
 
 //! Scenario21: Load two models, set them up to run chained after one another on the same gpu hardware device
-TEST_F(ScenarioCppWinrtGpuTest, DISABLED_Scenario21RunModel2ChainZ)
-{
-    // load a model, TODO: get a model that has an image descriptor
-    std::wstring filePath = FileHelpers::GetModulePath() + L"fns-candy.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create both session on the default gpu
-    LearningModelSession session1(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
-    LearningModelSession session2(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
-    // create both binding sets
-    LearningModelBinding binding1(session1);
-    LearningModelBinding binding2(session2);
-    // get the input descriptor
-    auto input = model.InputFeatures().GetAt(0);
-    // load a SoftwareBitmap
-    auto sb = FileHelpers::GetSoftwareBitmapFromFile(FileHelpers::GetModulePath() + L"fish_720.png");
-    auto videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
-    // bind it
-    binding1.Bind(input.Name(), videoFrame);
-    // get the output descriptor
-    auto output = model.OutputFeatures().GetAt(0);
-    // create an empty output tensor since we don't want the first model to detensorize into an image.
-
-    std::vector<int64_t> shape = { 1, 3, 720, 720 };
-    auto outputValue = TensorFloat::Create(shape);  //   FeatureValueFromFeatureValueDescriptor(input, nullptr);
-                                                    // now bind the(empty) output so we have a marker to chain with
-    binding1.Bind(output.Name(), outputValue);
-    // and leave the output unbound on the second model, we will fetch it later
-    // run both models async
-    EXPECT_NO_THROW(session1.EvaluateAsync(binding1, L""));
-
-    // now bind that output to the next models input
-    binding2.Bind(input.Name(), outputValue);
-
-    //eval the second model
-    auto session2AsyncOp = session2.EvaluateAsync(binding2, L"");
-
-    // now get the output don't wait, queue up the next model
-    auto finalOutput = session2AsyncOp.get().Outputs().First().Current().Value();
+static void Scenario21RunModel2ChainZ() {
+  // load a model, TODO: get a model that has an image descriptor
+  std::wstring filePath = FileHelpers::GetModulePath() + L"fns-candy.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create both session on the default gpu
+  LearningModelSession session1(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
+  LearningModelSession session2(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
+  // create both binding sets
+  LearningModelBinding binding1(session1);
+  LearningModelBinding binding2(session2);
+  // get the input descriptor
+  auto input = model.InputFeatures().GetAt(0);
+  // load a SoftwareBitmap
+  auto sb = FileHelpers::GetSoftwareBitmapFromFile(FileHelpers::GetModulePath() + L"fish_720.png");
+  auto videoFrame = VideoFrame::CreateWithSoftwareBitmap(sb);
+  // bind it
+  binding1.Bind(input.Name(), videoFrame);
+  // get the output descriptor
+  auto output = model.OutputFeatures().GetAt(0);
+  // create an empty output tensor since we don't want the first model to detensorize into an image.
+
+  std::vector<int64_t> shape = {1, 3, 720, 720};
+  auto outputValue = TensorFloat::Create(shape);  //   FeatureValueFromFeatureValueDescriptor(input, nullptr);
+                                                  // now bind the(empty) output so we have a marker to chain with
+  binding1.Bind(output.Name(), outputValue);
+  // and leave the output unbound on the second model, we will fetch it later
+  // run both models async
+  WINML_EXPECT_NO_THROW(session1.EvaluateAsync(binding1, L""));
+
+  // now bind that output to the next models input
+  binding2.Bind(input.Name(), outputValue);
+
+  //eval the second model
+  auto session2AsyncOp = session2.EvaluateAsync(binding2, L"");
+
+  // now get the output don't wait, queue up the next model
+  auto finalOutput = session2AsyncOp.get().Outputs().First().Current().Value();
 }
 
-bool VerifyHelper(ImageFeatureValue actual, ImageFeatureValue expected)
-{
-    auto softwareBitmapActual = actual.VideoFrame().SoftwareBitmap();
-    auto softwareBitmapExpected = expected.VideoFrame().SoftwareBitmap();
-    EXPECT_EQ(softwareBitmapActual.PixelHeight(), softwareBitmapExpected.PixelHeight());
-    EXPECT_EQ(softwareBitmapActual.PixelWidth(), softwareBitmapExpected.PixelWidth());
-    EXPECT_EQ(softwareBitmapActual.BitmapPixelFormat(), softwareBitmapExpected.BitmapPixelFormat());
-
-    // 4 means 4 channels
-    uint32_t size = 4 * softwareBitmapActual.PixelHeight() * softwareBitmapActual.PixelWidth();
-
-    winrt::Windows::Storage::Streams::Buffer actualOutputBuffer(size);
-    winrt::Windows::Storage::Streams::Buffer expectedOutputBuffer(size);
-
-    softwareBitmapActual.CopyToBuffer(actualOutputBuffer);
-    softwareBitmapExpected.CopyToBuffer(expectedOutputBuffer);
-
-    byte* actualBytes;
-    actualOutputBuffer.try_as<::Windows::Storage::Streams::IBufferByteAccess>()->Buffer(&actualBytes);
-    byte* expectedBytes;
-    expectedOutputBuffer.try_as<::Windows::Storage::Streams::IBufferByteAccess>()->Buffer(&expectedBytes);
-
-    byte* pActualByte = actualBytes;
-    byte* pExpectedByte = expectedBytes;
-
-    // hard code, might need to be modified later.
-    const float cMaxErrorRate = 0.06f;
-    byte epsilon = 20;
-
-    UINT errors = 0;
-    for (uint32_t i = 0; i < size; i++, pActualByte++, pExpectedByte++)
-    {
-        auto diff = (*pActualByte - *pExpectedByte);
-        if (diff > epsilon)
-        {
-            errors++;
-        }
-    }
-    std::cout << "total errors is " << errors << "/" << size << ", errors rate is " << (float)errors / size << "\n";
+bool VerifyHelper(ImageFeatureValue actual, ImageFeatureValue expected) {
+  auto softwareBitmapActual = actual.VideoFrame().SoftwareBitmap();
+  auto softwareBitmapExpected = expected.VideoFrame().SoftwareBitmap();
+  WINML_EXPECT_EQUAL(softwareBitmapActual.PixelHeight(), softwareBitmapExpected.PixelHeight());
+  WINML_EXPECT_EQUAL(softwareBitmapActual.PixelWidth(), softwareBitmapExpected.PixelWidth());
+  WINML_EXPECT_EQUAL(softwareBitmapActual.BitmapPixelFormat(), softwareBitmapExpected.BitmapPixelFormat());
 
-    return ((float)errors / size < cMaxErrorRate);
-}
+  // 4 means 4 channels
+  uint32_t size = 4 * softwareBitmapActual.PixelHeight() * softwareBitmapActual.PixelWidth();
 
-TEST_F(ScenarioCppWinrtTest, DISABLED_Scenario22ImageBindingAsCPUTensor)
-{
-    std::wstring modulePath = FileHelpers::GetModulePath();
-    std::wstring inputImagePath = modulePath + L"fish_720.png";
-    std::wstring bmImagePath = modulePath + L"bm_fish_720.jpg";
-    std::wstring modelPath = modulePath + L"fns-candy.onnx";
+  winrt::Windows::Storage::Streams::Buffer actualOutputBuffer(size);
+  winrt::Windows::Storage::Streams::Buffer expectedOutputBuffer(size);
 
-    auto device = LearningModelDevice(LearningModelDeviceKind::Default);
-    auto model = LearningModel::LoadFromFilePath(modelPath);
-    auto session = LearningModelSession(model, device);
-    auto binding = LearningModelBinding(session);
+  softwareBitmapActual.CopyToBuffer(actualOutputBuffer);
+  softwareBitmapExpected.CopyToBuffer(expectedOutputBuffer);
 
-    SoftwareBitmap softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(inputImagePath);
-    softwareBitmap = SoftwareBitmap::Convert(softwareBitmap, BitmapPixelFormat::Bgra8);
-
-    // Put softwareBitmap into buffer
-    BYTE* pData = nullptr;
-    UINT32 size = 0;
-    winrt::Windows::Graphics::Imaging::BitmapBuffer spBitmapBuffer(softwareBitmap.LockBuffer(winrt::Windows::Graphics::Imaging::BitmapBufferAccessMode::Read));
-    winrt::Windows::Foundation::IMemoryBufferReference reference = spBitmapBuffer.CreateReference();
-    auto spByteAccess = reference.as<::Windows::Foundation::IMemoryBufferByteAccess>();
-    spByteAccess->GetBuffer(&pData, &size);
-
-    std::vector<int64_t> shape = { 1, 3, softwareBitmap.PixelHeight() , softwareBitmap.PixelWidth() };
-    float* pCPUTensor;
-    uint32_t uCapacity;
-    TensorFloat tf = TensorFloat::Create(shape);
-    com_ptr<ITensorNative> itn = tf.as<ITensorNative>();
-    itn->GetBuffer(reinterpret_cast<BYTE**>(&pCPUTensor), &uCapacity);
-
-    uint32_t height = softwareBitmap.PixelHeight();
-    uint32_t width = softwareBitmap.PixelWidth();
-    for (UINT32 i = 0; i < size; i += 4)
-    {
-        UINT32 pixelInd = i / 4;
-        pCPUTensor[pixelInd] = (float)pData[i];
-        pCPUTensor[(height * width) + pixelInd] = (float)pData[i + 1];
-        pCPUTensor[(height * width * 2) + pixelInd] = (float)pData[i + 2];
-    }
+  byte* actualBytes;
+  actualOutputBuffer.try_as<::Windows::Storage::Streams::IBufferByteAccess>()->Buffer(&actualBytes);
+  byte* expectedBytes;
+  expectedOutputBuffer.try_as<::Windows::Storage::Streams::IBufferByteAccess>()->Buffer(&expectedBytes);
 
-    // Bind input
-    binding.Bind(model.InputFeatures().First().Current().Name(), tf);
-
-    // Bind output
-    auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
-    auto outputtensorshape = outputtensordescriptor.Shape();
-    VideoFrame outputimage(
-        BitmapPixelFormat::Bgra8,
-        static_cast<int32_t>(outputtensorshape.GetAt(3)),
-        static_cast<int32_t>(outputtensorshape.GetAt(2)));
-    ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
-    EXPECT_NO_THROW(binding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
-
-    // Evaluate the model
-    winrt::hstring correlationId;
-    EXPECT_NO_THROW(session.EvaluateAsync(binding, correlationId).get());
-
-    // Verify the output by comparing with the benchmark image
-    SoftwareBitmap bm_softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(bmImagePath);
-    bm_softwareBitmap = SoftwareBitmap::Convert(bm_softwareBitmap, BitmapPixelFormat::Bgra8);
-    VideoFrame bm_videoFrame = VideoFrame::CreateWithSoftwareBitmap(bm_softwareBitmap);
-    ImageFeatureValue bm_imagevalue = ImageFeatureValue::CreateFromVideoFrame(bm_videoFrame);
-    EXPECT_TRUE(VerifyHelper(bm_imagevalue, outputTensor));
-
-    // check the output video frame object by saving output image to disk
-    std::wstring outputDataImageFileName = L"out_cpu_tensor_fish_720.jpg";
-    StorageFolder currentfolder = StorageFolder::GetFolderFromPathAsync(modulePath).get();
-    StorageFile outimagefile = currentfolder.CreateFileAsync(outputDataImageFileName, CreationCollisionOption::ReplaceExisting).get();
-    IRandomAccessStream writestream = outimagefile.OpenAsync(FileAccessMode::ReadWrite).get();
-    BitmapEncoder encoder = BitmapEncoder::CreateAsync(BitmapEncoder::JpegEncoderId(), writestream).get();
-    // Set the software bitmap
-    encoder.SetSoftwareBitmap(outputimage.SoftwareBitmap());
-    encoder.FlushAsync().get();
-}
+  byte* pActualByte = actualBytes;
+  byte* pExpectedByte = expectedBytes;
+
+  // hard code, might need to be modified later.
+  const float cMaxErrorRate = 0.06f;
+  byte epsilon = 20;
 
-TEST_F(ScenarioCppWinrtGpuTest, DISABLED_Scenario22ImageBindingAsGPUTensor)
-{
-    std::wstring modulePath = FileHelpers::GetModulePath();
-    std::wstring inputImagePath = modulePath + L"fish_720.png";
-    std::wstring bmImagePath = modulePath + L"bm_fish_720.jpg";
-    std::wstring modelPath = modulePath + L"fns-candy.onnx";
-    std::wstring outputDataImageFileName = L"out_gpu_tensor_fish_720.jpg";
-
-    SoftwareBitmap softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(inputImagePath);
-    softwareBitmap = SoftwareBitmap::Convert(softwareBitmap, BitmapPixelFormat::Bgra8);
-
-    // Put softwareBitmap into cpu buffer
-    BYTE* pData = nullptr;
-    UINT32 size = 0;
-    winrt::Windows::Graphics::Imaging::BitmapBuffer spBitmapBuffer(softwareBitmap.LockBuffer(winrt::Windows::Graphics::Imaging::BitmapBufferAccessMode::Read));
-    winrt::Windows::Foundation::IMemoryBufferReference reference = spBitmapBuffer.CreateReference();
-    auto spByteAccess = reference.as<::Windows::Foundation::IMemoryBufferByteAccess>();
-    spByteAccess->GetBuffer(&pData, &size);
-
-    std::vector<int64_t> shape = { 1, 3, softwareBitmap.PixelHeight() , softwareBitmap.PixelWidth() };
-    FLOAT* pCPUTensor;
-    uint32_t uCapacity;
-
-    // CPU tensorization
-    TensorFloat tf = TensorFloat::Create(shape);
-    com_ptr<ITensorNative> itn = tf.as<ITensorNative>();
-    itn->GetBuffer(reinterpret_cast<BYTE**>(&pCPUTensor), &uCapacity);
-
-    uint32_t height = softwareBitmap.PixelHeight();
-    uint32_t width = softwareBitmap.PixelWidth();
-    for (UINT32 i = 0; i < size; i += 4)
-    {
-        UINT32 pixelInd = i / 4;
-        pCPUTensor[pixelInd] = (FLOAT)pData[i];
-        pCPUTensor[(height * width) + pixelInd] = (FLOAT)pData[i + 1];
-        pCPUTensor[(height * width * 2) + pixelInd] = (FLOAT)pData[i + 2];
+  UINT errors = 0;
+  for (uint32_t i = 0; i < size; i++, pActualByte++, pExpectedByte++) {
+    auto diff = (*pActualByte - *pExpectedByte);
+    if (diff > epsilon) {
+      errors++;
     }
+  }
+  std::cout << "total errors is " << errors << "/" << size << ", errors rate is " << (float)errors / size << "\n";
 
-    // create the d3d device.
-    com_ptr<ID3D12Device> pD3D12Device = nullptr;
-    EXPECT_NO_THROW(D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device), reinterpret_cast<void**>(&pD3D12Device)));
-
-    // create the command queue.
-    com_ptr<ID3D12CommandQueue> dxQueue = nullptr;
-    D3D12_COMMAND_QUEUE_DESC commandQueueDesc = {};
-    commandQueueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
-    pD3D12Device->CreateCommandQueue(&commandQueueDesc, __uuidof(ID3D12CommandQueue), reinterpret_cast<void**>(&dxQueue));
-    auto devicefactory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
-    auto tensorfactory = get_activation_factory<TensorFloat, ITensorStaticsNative>();
-    com_ptr<::IUnknown> spUnk;
-    devicefactory->CreateFromD3D12CommandQueue(dxQueue.get(), spUnk.put());
-
-    LearningModel model(nullptr);
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(modelPath));
-    LearningModelDevice dmlDeviceCustom = nullptr;
-    EXPECT_NO_THROW(spUnk.as(dmlDeviceCustom));
-    LearningModelSession dmlSessionCustom = nullptr;
-    EXPECT_NO_THROW(dmlSessionCustom = LearningModelSession(model, dmlDeviceCustom));
-    LearningModelBinding modelBinding = nullptr;
-    EXPECT_NO_THROW(modelBinding = LearningModelBinding(dmlSessionCustom));
-
-    // Create ID3D12GraphicsCommandList and Allocator
-    D3D12_COMMAND_LIST_TYPE queuetype = dxQueue->GetDesc().Type;
-    com_ptr<ID3D12CommandAllocator> alloctor;
-    com_ptr<ID3D12GraphicsCommandList> cmdList;
-
-    pD3D12Device->CreateCommandAllocator(
-        queuetype,
-        winrt::guid_of<ID3D12CommandAllocator>(),
-        alloctor.put_void());
-
-    pD3D12Device->CreateCommandList(
-        0,
-        queuetype,
-        alloctor.get(),
-        nullptr,
-        winrt::guid_of<ID3D12CommandList>(),
-        cmdList.put_void());
-
-    // Create Committed Resource
-    // 3 is number of channels we use. R G B without alpha.
-    UINT64 bufferbytesize = 3 * sizeof(float) * softwareBitmap.PixelWidth()*softwareBitmap.PixelHeight();
-    D3D12_HEAP_PROPERTIES heapProperties = {
-        D3D12_HEAP_TYPE_DEFAULT,
-        D3D12_CPU_PAGE_PROPERTY_UNKNOWN,
-        D3D12_MEMORY_POOL_UNKNOWN,
-        0,
-        0
-    };
-    D3D12_RESOURCE_DESC resourceDesc = {
-        D3D12_RESOURCE_DIMENSION_BUFFER,
-        0,
-        bufferbytesize,
-        1,
-        1,
-        1,
-        DXGI_FORMAT_UNKNOWN,
-    { 1, 0 },
-    D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
-    D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS
-    };
-
-    com_ptr<ID3D12Resource> pGPUResource = nullptr;
-    com_ptr<ID3D12Resource> imageUploadHeap;
-    pD3D12Device->CreateCommittedResource(
-        &heapProperties,
-        D3D12_HEAP_FLAG_NONE,
-        &resourceDesc,
-        D3D12_RESOURCE_STATE_COMMON,
-        nullptr,
-        __uuidof(ID3D12Resource),
-        pGPUResource.put_void()
-    );
-
-    // Create the GPU upload buffer.
-    CD3DX12_HEAP_PROPERTIES props(D3D12_HEAP_TYPE_UPLOAD);
-    auto buffer = CD3DX12_RESOURCE_DESC::Buffer(bufferbytesize);
-    EXPECT_NO_THROW(pD3D12Device->CreateCommittedResource(
-        &props,
-        D3D12_HEAP_FLAG_NONE,
-        &buffer,
-        D3D12_RESOURCE_STATE_GENERIC_READ,
-        nullptr,
-        __uuidof(ID3D12Resource),
-        imageUploadHeap.put_void()));
-
-    // Copy from Cpu to GPU
-    D3D12_SUBRESOURCE_DATA CPUData = {};
-    CPUData.pData = reinterpret_cast<BYTE*>(pCPUTensor);
-    CPUData.RowPitch = bufferbytesize;
-    CPUData.SlicePitch = bufferbytesize;
-    UpdateSubresources(cmdList.get(), pGPUResource.get(), imageUploadHeap.get(), 0, 0, 1, &CPUData);
-
-    // Close the command list and execute it to begin the initial GPU setup.
-    EXPECT_NO_THROW(cmdList->Close());
-    ID3D12CommandList* ppCommandLists[] = { cmdList.get() };
-    dxQueue->ExecuteCommandLists(_countof(ppCommandLists), ppCommandLists);
-
-    // GPU tensorize
-    com_ptr<::IUnknown> spUnkTensor;
-    TensorFloat input1imagetensor(nullptr);
-    __int64 shapes[4] = { 1,3, softwareBitmap.PixelWidth(), softwareBitmap.PixelHeight() };
-    tensorfactory->CreateFromD3D12Resource(pGPUResource.get(), shapes, 4, spUnkTensor.put());
-    spUnkTensor.try_as(input1imagetensor);
-
-    auto feature = model.InputFeatures().First();
-    EXPECT_NO_THROW(modelBinding.Bind(feature.Current().Name(), input1imagetensor));
-
-    auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
-    auto outputtensorshape = outputtensordescriptor.Shape();
-    VideoFrame outputimage(
-        BitmapPixelFormat::Rgba8,
-        static_cast<int32_t>(outputtensorshape.GetAt(3)),
-        static_cast<int32_t>(outputtensorshape.GetAt(2)));
-    ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
-
-    EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
-
-    // Evaluate the model
-    winrt::hstring correlationId;
-    dmlSessionCustom.EvaluateAsync(modelBinding, correlationId).get();
-
-    // Verify the output by comparing with the benchmark image
-    SoftwareBitmap bm_softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(bmImagePath);
-    bm_softwareBitmap = SoftwareBitmap::Convert(bm_softwareBitmap, BitmapPixelFormat::Rgba8);
-    VideoFrame bm_videoFrame = VideoFrame::CreateWithSoftwareBitmap(bm_softwareBitmap);
-    ImageFeatureValue bm_imagevalue = ImageFeatureValue::CreateFromVideoFrame(bm_videoFrame);
-    EXPECT_TRUE(VerifyHelper(bm_imagevalue, outputTensor));
-
-
-    //check the output video frame object
-    StorageFolder currentfolder = StorageFolder::GetFolderFromPathAsync(modulePath).get();
-    StorageFile outimagefile = currentfolder.CreateFileAsync(outputDataImageFileName, CreationCollisionOption::ReplaceExisting).get();
-    IRandomAccessStream writestream = outimagefile.OpenAsync(FileAccessMode::ReadWrite).get();
-    BitmapEncoder encoder = BitmapEncoder::CreateAsync(BitmapEncoder::JpegEncoderId(), writestream).get();
-    // Set the software bitmap
-    encoder.SetSoftwareBitmap(outputimage.SoftwareBitmap());
-    encoder.FlushAsync().get();
+  return ((float)errors / size < cMaxErrorRate);
 }
 
-TEST_F(ScenarioCppWinrtTest, QuantizedModels)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"onnxzoo_lotus_inception_v1-dq.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a session on the default device
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        binding.Bind(input.Name(), featureValue);
-    }
-    // run eval
-    EXPECT_NO_THROW(session.Evaluate(binding, filePath));
+static void Scenario22ImageBindingAsCPUTensor() {
+  std::wstring modulePath = FileHelpers::GetModulePath();
+  std::wstring inputImagePath = modulePath + L"fish_720.png";
+  std::wstring bmImagePath = modulePath + L"bm_fish_720.jpg";
+  std::wstring modelPath = modulePath + L"fns-candy.onnx";
+
+  auto device = LearningModelDevice(LearningModelDeviceKind::Default);
+  auto model = LearningModel::LoadFromFilePath(modelPath);
+  auto session = LearningModelSession(model, device);
+  auto binding = LearningModelBinding(session);
+
+  SoftwareBitmap softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(inputImagePath);
+  softwareBitmap = SoftwareBitmap::Convert(softwareBitmap, BitmapPixelFormat::Bgra8);
+
+  // Put softwareBitmap into buffer
+  BYTE* pData = nullptr;
+  UINT32 size = 0;
+  winrt::Windows::Graphics::Imaging::BitmapBuffer spBitmapBuffer(softwareBitmap.LockBuffer(winrt::Windows::Graphics::Imaging::BitmapBufferAccessMode::Read));
+  winrt::Windows::Foundation::IMemoryBufferReference reference = spBitmapBuffer.CreateReference();
+  auto spByteAccess = reference.as<::Windows::Foundation::IMemoryBufferByteAccess>();
+  spByteAccess->GetBuffer(&pData, &size);
+
+  std::vector<int64_t> shape = {1, 3, softwareBitmap.PixelHeight(), softwareBitmap.PixelWidth()};
+  float* pCPUTensor;
+  uint32_t uCapacity;
+  TensorFloat tf = TensorFloat::Create(shape);
+  com_ptr<ITensorNative> itn = tf.as<ITensorNative>();
+  itn->GetBuffer(reinterpret_cast<BYTE**>(&pCPUTensor), &uCapacity);
+
+  uint32_t height = softwareBitmap.PixelHeight();
+  uint32_t width = softwareBitmap.PixelWidth();
+  for (UINT32 i = 0; i < size; i += 4) {
+    UINT32 pixelInd = i / 4;
+    pCPUTensor[pixelInd] = (float)pData[i];
+    pCPUTensor[(height * width) + pixelInd] = (float)pData[i + 1];
+    pCPUTensor[(height * width * 2) + pixelInd] = (float)pData[i + 2];
+  }
+
+  // Bind input
+  binding.Bind(model.InputFeatures().First().Current().Name(), tf);
+
+  // Bind output
+  auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
+  auto outputtensorshape = outputtensordescriptor.Shape();
+  VideoFrame outputimage(
+      BitmapPixelFormat::Bgra8,
+      static_cast<int32_t>(outputtensorshape.GetAt(3)),
+      static_cast<int32_t>(outputtensorshape.GetAt(2)));
+  ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
+  WINML_EXPECT_NO_THROW(binding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
+
+  // Evaluate the model
+  winrt::hstring correlationId;
+  WINML_EXPECT_NO_THROW(session.EvaluateAsync(binding, correlationId).get());
+
+  // Verify the output by comparing with the benchmark image
+  SoftwareBitmap bm_softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(bmImagePath);
+  bm_softwareBitmap = SoftwareBitmap::Convert(bm_softwareBitmap, BitmapPixelFormat::Bgra8);
+  VideoFrame bm_videoFrame = VideoFrame::CreateWithSoftwareBitmap(bm_softwareBitmap);
+  ImageFeatureValue bm_imagevalue = ImageFeatureValue::CreateFromVideoFrame(bm_videoFrame);
+  WINML_EXPECT_TRUE(VerifyHelper(bm_imagevalue, outputTensor));
+
+  // check the output video frame object by saving output image to disk
+  std::wstring outputDataImageFileName = L"out_cpu_tensor_fish_720.jpg";
+  StorageFolder currentfolder = StorageFolder::GetFolderFromPathAsync(modulePath).get();
+  StorageFile outimagefile = currentfolder.CreateFileAsync(outputDataImageFileName, CreationCollisionOption::ReplaceExisting).get();
+  IRandomAccessStream writestream = outimagefile.OpenAsync(FileAccessMode::ReadWrite).get();
+  BitmapEncoder encoder = BitmapEncoder::CreateAsync(BitmapEncoder::JpegEncoderId(), writestream).get();
+  // Set the software bitmap
+  encoder.SetSoftwareBitmap(outputimage.SoftwareBitmap());
+  encoder.FlushAsync().get();
 }
 
-TEST_F(ScenarioCppWinrtGpuTest, MsftQuantizedModels)
-{
-    // load a model
-    std::wstring filePath = FileHelpers::GetModulePath() + L"coreml_Resnet50_ImageNet-dq.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
-    // create a binding set
-    LearningModelBinding binding(session);
-    // bind the input and the output buffers by name
-
-    std::wstring fullImagePath = FileHelpers::GetModulePath() + L"kitten_224.png";
-    StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
-    IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
-    SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
-
-    auto inputs = model.InputFeatures();
-    for (auto&& input : inputs)
-    {
-        auto featureValue = FeatureValueFromFeatureValueDescriptor(input, softwareBitmap);
-        // set an actual buffer here. we're using uninitialized data for simplicity.
-        binding.Bind(input.Name(), featureValue);
-    }
-    // run eval
-    EXPECT_NO_THROW(session.Evaluate(binding, filePath));
+static void Scenario22ImageBindingAsGPUTensor() {
+  std::wstring modulePath = FileHelpers::GetModulePath();
+  std::wstring inputImagePath = modulePath + L"fish_720.png";
+  std::wstring bmImagePath = modulePath + L"bm_fish_720.jpg";
+  std::wstring modelPath = modulePath + L"fns-candy.onnx";
+  std::wstring outputDataImageFileName = L"out_gpu_tensor_fish_720.jpg";
+
+  SoftwareBitmap softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(inputImagePath);
+  softwareBitmap = SoftwareBitmap::Convert(softwareBitmap, BitmapPixelFormat::Bgra8);
+
+  // Put softwareBitmap into cpu buffer
+  BYTE* pData = nullptr;
+  UINT32 size = 0;
+  winrt::Windows::Graphics::Imaging::BitmapBuffer spBitmapBuffer(softwareBitmap.LockBuffer(winrt::Windows::Graphics::Imaging::BitmapBufferAccessMode::Read));
+  winrt::Windows::Foundation::IMemoryBufferReference reference = spBitmapBuffer.CreateReference();
+  auto spByteAccess = reference.as<::Windows::Foundation::IMemoryBufferByteAccess>();
+  spByteAccess->GetBuffer(&pData, &size);
+
+  std::vector<int64_t> shape = {1, 3, softwareBitmap.PixelHeight(), softwareBitmap.PixelWidth()};
+  FLOAT* pCPUTensor;
+  uint32_t uCapacity;
+
+  // CPU tensorization
+  TensorFloat tf = TensorFloat::Create(shape);
+  com_ptr<ITensorNative> itn = tf.as<ITensorNative>();
+  itn->GetBuffer(reinterpret_cast<BYTE**>(&pCPUTensor), &uCapacity);
+
+  uint32_t height = softwareBitmap.PixelHeight();
+  uint32_t width = softwareBitmap.PixelWidth();
+  for (UINT32 i = 0; i < size; i += 4) {
+    UINT32 pixelInd = i / 4;
+    pCPUTensor[pixelInd] = (FLOAT)pData[i];
+    pCPUTensor[(height * width) + pixelInd] = (FLOAT)pData[i + 1];
+    pCPUTensor[(height * width * 2) + pixelInd] = (FLOAT)pData[i + 2];
+  }
+
+  // create the d3d device.
+  com_ptr<ID3D12Device> pD3D12Device = nullptr;
+  WINML_EXPECT_NO_THROW(D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL::D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device), reinterpret_cast<void**>(&pD3D12Device)));
+
+  // create the command queue.
+  com_ptr<ID3D12CommandQueue> dxQueue = nullptr;
+  D3D12_COMMAND_QUEUE_DESC commandQueueDesc = {};
+  commandQueueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
+  pD3D12Device->CreateCommandQueue(&commandQueueDesc, __uuidof(ID3D12CommandQueue), reinterpret_cast<void**>(&dxQueue));
+  auto devicefactory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
+  auto tensorfactory = get_activation_factory<TensorFloat, ITensorStaticsNative>();
+  com_ptr<::IUnknown> spUnk;
+  devicefactory->CreateFromD3D12CommandQueue(dxQueue.get(), spUnk.put());
+
+  LearningModel model(nullptr);
+  WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(modelPath));
+  LearningModelDevice dmlDeviceCustom = nullptr;
+  WINML_EXPECT_NO_THROW(spUnk.as(dmlDeviceCustom));
+  LearningModelSession dmlSessionCustom = nullptr;
+  WINML_EXPECT_NO_THROW(dmlSessionCustom = LearningModelSession(model, dmlDeviceCustom));
+  LearningModelBinding modelBinding = nullptr;
+  WINML_EXPECT_NO_THROW(modelBinding = LearningModelBinding(dmlSessionCustom));
+
+  // Create ID3D12GraphicsCommandList and Allocator
+  D3D12_COMMAND_LIST_TYPE queuetype = dxQueue->GetDesc().Type;
+  com_ptr<ID3D12CommandAllocator> alloctor;
+  com_ptr<ID3D12GraphicsCommandList> cmdList;
+
+  pD3D12Device->CreateCommandAllocator(
+      queuetype,
+      winrt::guid_of<ID3D12CommandAllocator>(),
+      alloctor.put_void());
+
+  pD3D12Device->CreateCommandList(
+      0,
+      queuetype,
+      alloctor.get(),
+      nullptr,
+      winrt::guid_of<ID3D12CommandList>(),
+      cmdList.put_void());
+
+  // Create Committed Resource
+  // 3 is number of channels we use. R G B without alpha.
+  UINT64 bufferbytesize = 3 * sizeof(float) * softwareBitmap.PixelWidth() * softwareBitmap.PixelHeight();
+  D3D12_HEAP_PROPERTIES heapProperties = {
+      D3D12_HEAP_TYPE_DEFAULT,
+      D3D12_CPU_PAGE_PROPERTY_UNKNOWN,
+      D3D12_MEMORY_POOL_UNKNOWN,
+      0,
+      0};
+  D3D12_RESOURCE_DESC resourceDesc = {
+      D3D12_RESOURCE_DIMENSION_BUFFER,
+      0,
+      bufferbytesize,
+      1,
+      1,
+      1,
+      DXGI_FORMAT_UNKNOWN,
+      {1, 0},
+      D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
+      D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS};
+
+  com_ptr<ID3D12Resource> pGPUResource = nullptr;
+  com_ptr<ID3D12Resource> imageUploadHeap;
+  pD3D12Device->CreateCommittedResource(
+      &heapProperties,
+      D3D12_HEAP_FLAG_NONE,
+      &resourceDesc,
+      D3D12_RESOURCE_STATE_COMMON,
+      nullptr,
+      __uuidof(ID3D12Resource),
+      pGPUResource.put_void());
+
+  // Create the GPU upload buffer.
+  CD3DX12_HEAP_PROPERTIES props(D3D12_HEAP_TYPE_UPLOAD);
+  auto buffer = CD3DX12_RESOURCE_DESC::Buffer(bufferbytesize);
+  WINML_EXPECT_NO_THROW(pD3D12Device->CreateCommittedResource(
+      &props,
+      D3D12_HEAP_FLAG_NONE,
+      &buffer,
+      D3D12_RESOURCE_STATE_GENERIC_READ,
+      nullptr,
+      __uuidof(ID3D12Resource),
+      imageUploadHeap.put_void()));
+
+  // Copy from Cpu to GPU
+  D3D12_SUBRESOURCE_DATA CPUData = {};
+  CPUData.pData = reinterpret_cast<BYTE*>(pCPUTensor);
+  CPUData.RowPitch = bufferbytesize;
+  CPUData.SlicePitch = bufferbytesize;
+  UpdateSubresources(cmdList.get(), pGPUResource.get(), imageUploadHeap.get(), 0, 0, 1, &CPUData);
+
+  // Close the command list and execute it to begin the initial GPU setup.
+  WINML_EXPECT_NO_THROW(cmdList->Close());
+  ID3D12CommandList* ppCommandLists[] = {cmdList.get()};
+  dxQueue->ExecuteCommandLists(_countof(ppCommandLists), ppCommandLists);
+
+  // GPU tensorize
+  com_ptr<::IUnknown> spUnkTensor;
+  TensorFloat input1imagetensor(nullptr);
+  __int64 shapes[4] = {1, 3, softwareBitmap.PixelWidth(), softwareBitmap.PixelHeight()};
+  tensorfactory->CreateFromD3D12Resource(pGPUResource.get(), shapes, 4, spUnkTensor.put());
+  spUnkTensor.try_as(input1imagetensor);
+
+  auto feature = model.InputFeatures().First();
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(feature.Current().Name(), input1imagetensor));
+
+  auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
+  auto outputtensorshape = outputtensordescriptor.Shape();
+  VideoFrame outputimage(
+      BitmapPixelFormat::Rgba8,
+      static_cast<int32_t>(outputtensorshape.GetAt(3)),
+      static_cast<int32_t>(outputtensorshape.GetAt(2)));
+  ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
+
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
+
+  // Evaluate the model
+  winrt::hstring correlationId;
+  dmlSessionCustom.EvaluateAsync(modelBinding, correlationId).get();
+
+  // Verify the output by comparing with the benchmark image
+  SoftwareBitmap bm_softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(bmImagePath);
+  bm_softwareBitmap = SoftwareBitmap::Convert(bm_softwareBitmap, BitmapPixelFormat::Rgba8);
+  VideoFrame bm_videoFrame = VideoFrame::CreateWithSoftwareBitmap(bm_softwareBitmap);
+  ImageFeatureValue bm_imagevalue = ImageFeatureValue::CreateFromVideoFrame(bm_videoFrame);
+  WINML_EXPECT_TRUE(VerifyHelper(bm_imagevalue, outputTensor));
+
+  //check the output video frame object
+  StorageFolder currentfolder = StorageFolder::GetFolderFromPathAsync(modulePath).get();
+  StorageFile outimagefile = currentfolder.CreateFileAsync(outputDataImageFileName, CreationCollisionOption::ReplaceExisting).get();
+  IRandomAccessStream writestream = outimagefile.OpenAsync(FileAccessMode::ReadWrite).get();
+  BitmapEncoder encoder = BitmapEncoder::CreateAsync(BitmapEncoder::JpegEncoderId(), writestream).get();
+  // Set the software bitmap
+  encoder.SetSoftwareBitmap(outputimage.SoftwareBitmap());
+  encoder.FlushAsync().get();
 }
 
-TEST_F(ScenarioCppWinrtGpuTest, DISABLED_SyncVsAsync)
-{
-    // create model, device and session
-    LearningModel model = nullptr;
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(FileHelpers::GetModulePath() + L"fns-candy.onnx"));
-
-    LearningModelSession session = nullptr;
-    EXPECT_NO_THROW(session = LearningModelSession(model, LearningModelDevice(LearningModelDeviceKind::DirectX)));
-
-    // create the binding
-    LearningModelBinding modelBinding(session);
-
-    // bind the input
-    std::wstring fullImagePath = FileHelpers::GetModulePath() + L"fish_720.png";
-    StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
-    IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
-    SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
-    VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
-
-    auto imagetensor = ImageFeatureValue::CreateFromVideoFrame(frame);
-    auto inputFeatureDescriptor = model.InputFeatures().First();
-    EXPECT_NO_THROW(modelBinding.Bind(inputFeatureDescriptor.Current().Name(), imagetensor));
-
-    UINT N = 20;
-
-    auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
-    auto outputtensorshape = outputtensordescriptor.Shape();
-    VideoFrame outputimage(
-        BitmapPixelFormat::Rgba8,
-        static_cast<int32_t>(outputtensorshape.GetAt(3)),
-        static_cast<int32_t>(outputtensorshape.GetAt(2)));
-    ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
-    EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
-
-    // evaluate N times synchronously and time it
-    auto startSync = std::chrono::high_resolution_clock::now();
-    for (UINT i = 0; i < N; i++)
-    {
-        session.Evaluate(modelBinding, L"");
-    }
-    auto syncTime = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::high_resolution_clock::now() - startSync);
-    std::cout << "Synchronous time for " << N << " evaluations: " << syncTime.count() << " milliseconds\n";
-
-    // evaluate N times Asynchronously and time it
-    std::vector<winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult>> tasks;
-    std::vector<LearningModelBinding> bindings(N, nullptr);
-
-    for (size_t i = 0; i < bindings.size(); i++)
-    {
-        bindings[i] = LearningModelBinding(session);
-        bindings[i].Bind(inputFeatureDescriptor.Current().Name(), imagetensor);
-        bindings[i].Bind(
-            model.OutputFeatures().First().Current().Name(),
-			VideoFrame(BitmapPixelFormat::Rgba8,
-				static_cast<int32_t>(outputtensorshape.GetAt(3)),
-				static_cast<int32_t>(outputtensorshape.GetAt(2))));
-    }
+static void QuantizedModels() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"onnxzoo_lotus_inception_v1-dq.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a session on the default device
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::Default));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    binding.Bind(input.Name(), featureValue);
+  }
+  // run eval
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, filePath));
+}
 
-    auto startAsync = std::chrono::high_resolution_clock::now();
-    for (UINT i = 0; i < N; i++)
-    {
-        tasks.emplace_back(session.EvaluateAsync(bindings[i], L""));
-    }
-    // wait for them all to complete
-    for (auto&& task : tasks)
-    {
-        task.get();
-    }
-    auto asyncTime = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::high_resolution_clock::now() - startAsync);
-    std::cout << "Asynchronous time for " << N << " evaluations: " << asyncTime.count() << " milliseconds\n";
+static void MsftQuantizedModels() {
+  // load a model
+  std::wstring filePath = FileHelpers::GetModulePath() + L"coreml_Resnet50_ImageNet-dq.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  LearningModelSession session(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
+  // create a binding set
+  LearningModelBinding binding(session);
+  // bind the input and the output buffers by name
+
+  std::wstring fullImagePath = FileHelpers::GetModulePath() + L"kitten_224.png";
+  StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
+  IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
+  SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
+
+  auto inputs = model.InputFeatures();
+  for (auto&& input : inputs) {
+    auto featureValue = FeatureValueFromFeatureValueDescriptor(input, softwareBitmap);
+    // set an actual buffer here. we're using uninitialized data for simplicity.
+    binding.Bind(input.Name(), featureValue);
+  }
+  // run eval
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, filePath));
 }
 
+static void SyncVsAsync() {
+  // create model, device and session
+  LearningModel model = nullptr;
+  WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(FileHelpers::GetModulePath() + L"fns-candy.onnx"));
+
+  LearningModelSession session = nullptr;
+  WINML_EXPECT_NO_THROW(session = LearningModelSession(model, LearningModelDevice(LearningModelDeviceKind::DirectX)));
+
+  // create the binding
+  LearningModelBinding modelBinding(session);
+
+  // bind the input
+  std::wstring fullImagePath = FileHelpers::GetModulePath() + L"fish_720.png";
+  StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
+  IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
+  SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
+  VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
+
+  auto imagetensor = ImageFeatureValue::CreateFromVideoFrame(frame);
+  auto inputFeatureDescriptor = model.InputFeatures().First();
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(inputFeatureDescriptor.Current().Name(), imagetensor));
+
+  UINT N = 20;
+
+  auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
+  auto outputtensorshape = outputtensordescriptor.Shape();
+  VideoFrame outputimage(
+      BitmapPixelFormat::Rgba8,
+      static_cast<int32_t>(outputtensorshape.GetAt(3)),
+      static_cast<int32_t>(outputtensorshape.GetAt(2)));
+  ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
+
+  // evaluate N times synchronously and time it
+  auto startSync = std::chrono::high_resolution_clock::now();
+  for (UINT i = 0; i < N; i++) {
+    session.Evaluate(modelBinding, L"");
+  }
+  auto syncTime = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::high_resolution_clock::now() - startSync);
+  std::cout << "Synchronous time for " << N << " evaluations: " << syncTime.count() << " milliseconds\n";
+
+  // evaluate N times Asynchronously and time it
+  std::vector<winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult>> tasks;
+  std::vector<LearningModelBinding> bindings(N, nullptr);
+
+  for (size_t i = 0; i < bindings.size(); i++) {
+    bindings[i] = LearningModelBinding(session);
+    bindings[i].Bind(inputFeatureDescriptor.Current().Name(), imagetensor);
+    bindings[i].Bind(
+        model.OutputFeatures().First().Current().Name(),
+        VideoFrame(BitmapPixelFormat::Rgba8,
+                   static_cast<int32_t>(outputtensorshape.GetAt(3)),
+                   static_cast<int32_t>(outputtensorshape.GetAt(2))));
+  }
 
-TEST_F(ScenarioCppWinrtGpuTest, DISABLED_CustomCommandQueueWithFence)
-{
-    static const wchar_t* const modelFileName = L"fns-candy.onnx";
-    static const wchar_t* const inputDataImageFileName = L"fish_720.png";
+  auto startAsync = std::chrono::high_resolution_clock::now();
+  for (UINT i = 0; i < N; i++) {
+    tasks.emplace_back(session.EvaluateAsync(bindings[i], L""));
+  }
+  // wait for them all to complete
+  for (auto&& task : tasks) {
+    task.get();
+  }
+  auto asyncTime = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::high_resolution_clock::now() - startAsync);
+  std::cout << "Asynchronous time for " << N << " evaluations: " << asyncTime.count() << " milliseconds\n";
+}
 
-    com_ptr<ID3D12Device> d3d12Device;
-    EXPECT_HRESULT_SUCCEEDED(D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device), d3d12Device.put_void()));
+static void CustomCommandQueueWithFence() {
+  static const wchar_t* const modelFileName = L"fns-candy.onnx";
+  static const wchar_t* const inputDataImageFileName = L"fish_720.png";
 
-    D3D12_COMMAND_QUEUE_DESC queueDesc = {};
-    queueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
+  com_ptr<ID3D12Device> d3d12Device;
+  WINML_EXPECT_HRESULT_SUCCEEDED(D3D12CreateDevice(nullptr, D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device), d3d12Device.put_void()));
 
-    com_ptr<ID3D12CommandQueue> queue;
-    EXPECT_HRESULT_SUCCEEDED(d3d12Device->CreateCommandQueue(&queueDesc, __uuidof(ID3D12CommandQueue), queue.put_void()));
+  D3D12_COMMAND_QUEUE_DESC queueDesc = {};
+  queueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
 
-    com_ptr<ID3D12Fence> fence;
-    EXPECT_HRESULT_SUCCEEDED(d3d12Device->CreateFence(0, D3D12_FENCE_FLAG_NONE, __uuidof(ID3D12Fence), fence.put_void()));
+  com_ptr<ID3D12CommandQueue> queue;
+  WINML_EXPECT_HRESULT_SUCCEEDED(d3d12Device->CreateCommandQueue(&queueDesc, __uuidof(ID3D12CommandQueue), queue.put_void()));
 
-    auto devicefactory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
+  com_ptr<ID3D12Fence> fence;
+  WINML_EXPECT_HRESULT_SUCCEEDED(d3d12Device->CreateFence(0, D3D12_FENCE_FLAG_NONE, __uuidof(ID3D12Fence), fence.put_void()));
 
-    com_ptr<::IUnknown> learningModelDeviceUnknown;
-    EXPECT_HRESULT_SUCCEEDED(devicefactory->CreateFromD3D12CommandQueue(queue.get(), learningModelDeviceUnknown.put()));
+  auto devicefactory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
 
-    LearningModelDevice device = nullptr;
-    EXPECT_NO_THROW(learningModelDeviceUnknown.as(device));
+  com_ptr<::IUnknown> learningModelDeviceUnknown;
+  WINML_EXPECT_HRESULT_SUCCEEDED(devicefactory->CreateFromD3D12CommandQueue(queue.get(), learningModelDeviceUnknown.put()));
 
-    std::wstring modulePath = FileHelpers::GetModulePath();
+  LearningModelDevice device = nullptr;
+  WINML_EXPECT_NO_THROW(learningModelDeviceUnknown.as(device));
 
-    // WinML model creation
-    std::wstring fullModelPath = modulePath + modelFileName;
-    LearningModel model(nullptr);
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
-    LearningModelSession modelSession = nullptr;
-    EXPECT_NO_THROW(modelSession = LearningModelSession(model, device));
-    LearningModelBinding modelBinding = nullptr;
-    EXPECT_NO_THROW(modelBinding = LearningModelBinding(modelSession));
+  std::wstring modulePath = FileHelpers::GetModulePath();
 
-    std::wstring fullImagePath = modulePath + inputDataImageFileName;
+  // WinML model creation
+  std::wstring fullModelPath = modulePath + modelFileName;
+  LearningModel model(nullptr);
+  WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
+  LearningModelSession modelSession = nullptr;
+  WINML_EXPECT_NO_THROW(modelSession = LearningModelSession(model, device));
+  LearningModelBinding modelBinding = nullptr;
+  WINML_EXPECT_NO_THROW(modelBinding = LearningModelBinding(modelSession));
 
-    StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
-    IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
-    SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
-    VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
-    ImageFeatureValue input1imagetensor = ImageFeatureValue::CreateFromVideoFrame(frame);
+  std::wstring fullImagePath = modulePath + inputDataImageFileName;
 
-    auto feature = model.InputFeatures().First();
-    EXPECT_NO_THROW(modelBinding.Bind(feature.Current().Name(), input1imagetensor));
+  StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
+  IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
+  SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
+  VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
+  ImageFeatureValue input1imagetensor = ImageFeatureValue::CreateFromVideoFrame(frame);
 
-    auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
-    auto outputtensorshape = outputtensordescriptor.Shape();
-    VideoFrame outputimage(
-        BitmapPixelFormat::Rgba8,
-        static_cast<int32_t>(outputtensorshape.GetAt(3)),
-        static_cast<int32_t>(outputtensorshape.GetAt(2)));
-    ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
+  auto feature = model.InputFeatures().First();
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(feature.Current().Name(), input1imagetensor));
 
-    EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
+  auto outputtensordescriptor = model.OutputFeatures().First().Current().as<ITensorFeatureDescriptor>();
+  auto outputtensorshape = outputtensordescriptor.Shape();
+  VideoFrame outputimage(
+      BitmapPixelFormat::Rgba8,
+      static_cast<int32_t>(outputtensorshape.GetAt(3)),
+      static_cast<int32_t>(outputtensorshape.GetAt(2)));
+  ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
 
-    // Block the queue on the fence, evaluate the model, then queue a signal. The model evaluation should not complete
-    // until after the wait is unblocked, and the signal should not complete until model evaluation does. This can
-    // only be true if WinML executes the workload on the supplied queue (instead of using its own).
+  WINML_EXPECT_NO_THROW(modelBinding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
 
-    EXPECT_HRESULT_SUCCEEDED(queue->Wait(fence.get(), 1));
+  // Block the queue on the fence, evaluate the model, then queue a signal. The model evaluation should not complete
+  // until after the wait is unblocked, and the signal should not complete until model evaluation does. This can
+  // only be true if WinML executes the workload on the supplied queue (instead of using its own).
 
-    EXPECT_HRESULT_SUCCEEDED(queue->Signal(fence.get(), 2));
+  WINML_EXPECT_HRESULT_SUCCEEDED(queue->Wait(fence.get(), 1));
 
-    winrt::hstring correlationId;
-    winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult> asyncOp;
-    EXPECT_NO_THROW(asyncOp = modelSession.EvaluateAsync(modelBinding, correlationId));
+  WINML_EXPECT_HRESULT_SUCCEEDED(queue->Signal(fence.get(), 2));
 
-    Sleep(1000); // Give the model a chance to run (which it shouldn't if everything is working correctly)
+  winrt::hstring correlationId;
+  winrt::Windows::Foundation::IAsyncOperation<LearningModelEvaluationResult> asyncOp;
+  WINML_EXPECT_NO_THROW(asyncOp = modelSession.EvaluateAsync(modelBinding, correlationId));
 
-    // Because we haven't unblocked the wait yet, model evaluation must not have completed (nor the fence signal)
-    EXPECT_NE(asyncOp.Status(), winrt::Windows::Foundation::AsyncStatus::Completed);
-    EXPECT_EQ(fence->GetCompletedValue(), 0);
+  Sleep(1000);  // Give the model a chance to run (which it shouldn't if everything is working correctly)
 
-    // Unblock the queue
-    EXPECT_HRESULT_SUCCEEDED(fence->Signal(1));
+  // Because we haven't unblocked the wait yet, model evaluation must not have completed (nor the fence signal)
+  WINML_EXPECT_NOT_EQUAL(asyncOp.Status(), winrt::Windows::Foundation::AsyncStatus::Completed);
+  WINML_EXPECT_EQUAL(fence->GetCompletedValue(), 0);
 
-    // Wait for model evaluation to complete
-    asyncOp.get();
+  // Unblock the queue
+  WINML_EXPECT_HRESULT_SUCCEEDED(fence->Signal(1));
 
-    // The fence must be signaled by now (because model evaluation has completed)
-    EXPECT_EQ(fence->GetCompletedValue(), 2);
-}
+  // Wait for model evaluation to complete
+  asyncOp.get();
 
-TEST_F(ScenarioCppWinrtGpuTest, DISABLED_ReuseVideoFrame)
-{
-    std::wstring modulePath = FileHelpers::GetModulePath();
-    std::wstring inputImagePath = modulePath + L"fish_720.png";
-    std::wstring bmImagePath = modulePath + L"bm_fish_720.jpg";
-    std::wstring modelPath = modulePath + L"fns-candy.onnx";
-
-    std::vector<LearningModelDeviceKind> deviceKinds = { LearningModelDeviceKind::Cpu, LearningModelDeviceKind::DirectX };
-    std::vector<std::string> videoFrameSources;
-    DeviceHelpers::AdapterEnumerationSupport support;
-    DeviceHelpers::GetAdapterEnumerationSupport(&support);
-    if (support.has_dxgi)
-    {
-        videoFrameSources = { "SoftwareBitmap", "Direct3DSurface" };
-    }
-    else {
-        videoFrameSources = { "SoftwareBitmap" };
+  // The fence must be signaled by now (because model evaluation has completed)
+  WINML_EXPECT_EQUAL(fence->GetCompletedValue(), 2);
+}
 
-    }
+static void ReuseVideoFrame() {
+  std::wstring modulePath = FileHelpers::GetModulePath();
+  std::wstring inputImagePath = modulePath + L"fish_720.png";
+  std::wstring bmImagePath = modulePath + L"bm_fish_720.jpg";
+  std::wstring modelPath = modulePath + L"fns-candy.onnx";
+
+  std::vector<LearningModelDeviceKind> deviceKinds = {LearningModelDeviceKind::Cpu, LearningModelDeviceKind::DirectX};
+  std::vector<std::string> videoFrameSources;
+  DeviceHelpers::AdapterEnumerationSupport support;
+  DeviceHelpers::GetAdapterEnumerationSupport(&support);
+  if (support.has_dxgi) {
+    videoFrameSources = {"SoftwareBitmap", "Direct3DSurface"};
+  } else {
+    videoFrameSources = {"SoftwareBitmap"};
+  }
 
-    for (auto deviceKind : deviceKinds)
-    {
-        auto device = LearningModelDevice(deviceKind);
-        auto model = LearningModel::LoadFromFilePath(modelPath);
-        auto session = LearningModelSession(model, device);
-        auto binding = LearningModelBinding(session);
-        for (auto videoFrameSource : videoFrameSources)
-        {
-            VideoFrame reuseVideoFrame = nullptr;
-            if (videoFrameSource == "SoftwareBitmap")
-            {
-                reuseVideoFrame = VideoFrame::CreateWithSoftwareBitmap(SoftwareBitmap(BitmapPixelFormat::Bgra8, 720, 720));
-            }
-            else
-            {
-                reuseVideoFrame = VideoFrame::CreateAsDirect3D11SurfaceBacked(DirectXPixelFormat::B8G8R8X8UIntNormalized, 720, 720);
-            }
-            for (uint32_t i = 0; i < 3; ++i)
-            {
-                SoftwareBitmap softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(inputImagePath);
-                VideoFrame videoFrame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
-                // reuse video frame
-                videoFrame.CopyToAsync(reuseVideoFrame).get();
-
-                // bind input
-                binding.Bind(model.InputFeatures().First().Current().Name(), reuseVideoFrame);
-
-                // bind output
-                VideoFrame outputimage(BitmapPixelFormat::Bgra8, 720, 720);
-                ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
-                EXPECT_NO_THROW(binding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
-
-                // evaluate
-                winrt::hstring correlationId;
-                EXPECT_NO_THROW(session.EvaluateAsync(binding, correlationId).get());
-
-                // verify result
-                SoftwareBitmap bm_softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(bmImagePath);
-                bm_softwareBitmap = SoftwareBitmap::Convert(bm_softwareBitmap, BitmapPixelFormat::Bgra8);
-                VideoFrame bm_videoFrame = VideoFrame::CreateWithSoftwareBitmap(bm_softwareBitmap);
-                ImageFeatureValue bm_imagevalue = ImageFeatureValue::CreateFromVideoFrame(bm_videoFrame);
-                EXPECT_TRUE(VerifyHelper(bm_imagevalue, outputTensor));
-            }
-        }
+  for (auto deviceKind : deviceKinds) {
+    auto device = LearningModelDevice(deviceKind);
+    auto model = LearningModel::LoadFromFilePath(modelPath);
+    auto session = LearningModelSession(model, device);
+    auto binding = LearningModelBinding(session);
+    for (auto videoFrameSource : videoFrameSources) {
+      VideoFrame reuseVideoFrame = nullptr;
+      if (videoFrameSource == "SoftwareBitmap") {
+        reuseVideoFrame = VideoFrame::CreateWithSoftwareBitmap(SoftwareBitmap(BitmapPixelFormat::Bgra8, 720, 720));
+      } else {
+        reuseVideoFrame = VideoFrame::CreateAsDirect3D11SurfaceBacked(DirectXPixelFormat::B8G8R8X8UIntNormalized, 720, 720);
+      }
+      for (uint32_t i = 0; i < 3; ++i) {
+        SoftwareBitmap softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(inputImagePath);
+        VideoFrame videoFrame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
+        // reuse video frame
+        videoFrame.CopyToAsync(reuseVideoFrame).get();
+
+        // bind input
+        binding.Bind(model.InputFeatures().First().Current().Name(), reuseVideoFrame);
+
+        // bind output
+        VideoFrame outputimage(BitmapPixelFormat::Bgra8, 720, 720);
+        ImageFeatureValue outputTensor = ImageFeatureValue::CreateFromVideoFrame(outputimage);
+        WINML_EXPECT_NO_THROW(binding.Bind(model.OutputFeatures().First().Current().Name(), outputTensor));
+
+        // evaluate
+        winrt::hstring correlationId;
+        WINML_EXPECT_NO_THROW(session.EvaluateAsync(binding, correlationId).get());
+
+        // verify result
+        SoftwareBitmap bm_softwareBitmap = FileHelpers::GetSoftwareBitmapFromFile(bmImagePath);
+        bm_softwareBitmap = SoftwareBitmap::Convert(bm_softwareBitmap, BitmapPixelFormat::Bgra8);
+        VideoFrame bm_videoFrame = VideoFrame::CreateWithSoftwareBitmap(bm_softwareBitmap);
+        ImageFeatureValue bm_imagevalue = ImageFeatureValue::CreateFromVideoFrame(bm_videoFrame);
+        WINML_EXPECT_TRUE(VerifyHelper(bm_imagevalue, outputTensor));
+      }
     }
+  }
 }
-
-TEST_F(ScenarioCppWinrtTest, EncryptedStream)
-{
-    // get a stream
-    std::wstring path = FileHelpers::GetModulePath() + L"model.onnx";
-    auto storageFile = StorageFile::GetFileFromPathAsync(path).get();
-    auto fileBuffer = winrt::Windows::Storage::FileIO::ReadBufferAsync(storageFile).get();
-
-    // encrypt
-    auto algorithmName = winrt::Windows::Security::Cryptography::Core::SymmetricAlgorithmNames::AesCbcPkcs7();
-    auto algorithm = winrt::Windows::Security::Cryptography::Core::SymmetricKeyAlgorithmProvider::OpenAlgorithm(algorithmName);
-    uint32_t keyLength = 32;
-    auto keyBuffer = winrt::Windows::Security::Cryptography::CryptographicBuffer::GenerateRandom(keyLength);
-    auto key = algorithm.CreateSymmetricKey(keyBuffer);
-    auto iv = winrt::Windows::Security::Cryptography::CryptographicBuffer::GenerateRandom(algorithm.BlockLength());
-    auto encryptedBuffer = winrt::Windows::Security::Cryptography::Core::CryptographicEngine::Encrypt(key, fileBuffer, iv);
-
-    // verify loading the encrypted stream fails appropriately.
-    auto encryptedStream = InMemoryRandomAccessStream();
-    encryptedStream.WriteAsync(encryptedBuffer).get();
-    EXPECT_THROW_SPECIFIC(LearningModel::LoadFromStream(RandomAccessStreamReference::CreateFromStream(encryptedStream)),
-        winrt::hresult_error,
-        [](const winrt::hresult_error& e) -> bool
-        {
-            return e.code() == E_INVALIDARG;
-        });
-
-    // now decrypt
-    auto decryptedBuffer = winrt::Windows::Security::Cryptography::Core::CryptographicEngine::Decrypt(key, encryptedBuffer, iv);
-    auto decryptedStream = InMemoryRandomAccessStream();
-    decryptedStream.WriteAsync(decryptedBuffer).get();
-
-    // load!
-    LearningModel model = nullptr;
-    EXPECT_NO_THROW(model = LearningModel::LoadFromStream(RandomAccessStreamReference::CreateFromStream(decryptedStream)));
-    LearningModelSession session = nullptr;
-    EXPECT_NO_THROW(session = LearningModelSession(model));
+static void EncryptedStream() {
+  // get a stream
+  std::wstring path = FileHelpers::GetModulePath() + L"model.onnx";
+  auto storageFile = StorageFile::GetFileFromPathAsync(path).get();
+  auto fileBuffer = winrt::Windows::Storage::FileIO::ReadBufferAsync(storageFile).get();
+
+  // encrypt
+  auto algorithmName = winrt::Windows::Security::Cryptography::Core::SymmetricAlgorithmNames::AesCbcPkcs7();
+  auto algorithm = winrt::Windows::Security::Cryptography::Core::SymmetricKeyAlgorithmProvider::OpenAlgorithm(algorithmName);
+  uint32_t keyLength = 32;
+  auto keyBuffer = winrt::Windows::Security::Cryptography::CryptographicBuffer::GenerateRandom(keyLength);
+  auto key = algorithm.CreateSymmetricKey(keyBuffer);
+  auto iv = winrt::Windows::Security::Cryptography::CryptographicBuffer::GenerateRandom(algorithm.BlockLength());
+  auto encryptedBuffer = winrt::Windows::Security::Cryptography::Core::CryptographicEngine::Encrypt(key, fileBuffer, iv);
+
+  // verify loading the encrypted stream fails appropriately.
+  auto encryptedStream = InMemoryRandomAccessStream();
+  encryptedStream.WriteAsync(encryptedBuffer).get();
+  WINML_EXPECT_THROW_SPECIFIC(LearningModel::LoadFromStream(RandomAccessStreamReference::CreateFromStream(encryptedStream)),
+                              winrt::hresult_error,
+                              [](const winrt::hresult_error& e) -> bool {
+                                return e.code() == E_INVALIDARG;
+                              });
+
+  // now decrypt
+  auto decryptedBuffer = winrt::Windows::Security::Cryptography::Core::CryptographicEngine::Decrypt(key, encryptedBuffer, iv);
+  auto decryptedStream = InMemoryRandomAccessStream();
+  decryptedStream.WriteAsync(decryptedBuffer).get();
+
+  // load!
+  LearningModel model = nullptr;
+  WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromStream(RandomAccessStreamReference::CreateFromStream(decryptedStream)));
+  LearningModelSession session = nullptr;
+  WINML_EXPECT_NO_THROW(session = LearningModelSession(model));
 }
 
-void DeviceLostRecoveryHelper()
-{
+static void DeviceLostRecovery() {
   // load a model
   std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
   LearningModel model = LearningModel::LoadFromFilePath(filePath);
@@ -1469,13 +1345,10 @@ void DeviceLostRecoveryHelper()
   }
 
   // evaluate should fail
-  try
-  {
+  try {
     session.Evaluate(binding, L"");
-    FAIL() << "Evaluate should fail after removing the device";
-  }
-  catch (...)
-  {
+    WINML_LOG_ERROR("Evaluate should fail after removing the device");
+  } catch (...) {
   }
 
   // remove all references to the device by reseting the session and binding.
@@ -1483,67 +1356,97 @@ void DeviceLostRecoveryHelper()
   binding = nullptr;
 
   // create new session and binding and try again!
-  session = LearningModelSession(model, LearningModelDevice(LearningModelDeviceKind::DirectX));
-  binding = LearningModelBinding(session);
+  WINML_EXPECT_NO_THROW(session = LearningModelSession(model, LearningModelDevice(LearningModelDeviceKind::DirectX)));
+  WINML_EXPECT_NO_THROW(binding = LearningModelBinding(session));
   BindFeatures(binding, model.InputFeatures());
-  session.Evaluate(binding, L"");
-  exit(0);
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
-TEST_F(ScenarioCppWinrtGpuTestDeathTest, DeviceLostRecovery) {
-
-    ::testing::FLAGS_gtest_death_test_style = "threadsafe";
-    EXPECT_EXIT(
-        DeviceLostRecoveryHelper(),
-        ::testing::ExitedWithCode(0),
-        ""
-    );
+static void D2DInterop() {
+  // load a model (model.onnx == squeezenet[1,3,224,224])
+  std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
+  LearningModel model = LearningModel::LoadFromFilePath(filePath);
+  // create a dx12 device
+  com_ptr<ID3D12Device1> device = nullptr;
+  WINML_EXPECT_HRESULT_SUCCEEDED(D3D12CreateDevice(NULL, D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device1), device.put_void()));
+  // now create a command queue from it
+  com_ptr<ID3D12CommandQueue> commandQueue = nullptr;
+  D3D12_COMMAND_QUEUE_DESC queueDesc = {};
+  queueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
+  WINML_EXPECT_HRESULT_SUCCEEDED(device->CreateCommandQueue(&queueDesc, winrt::guid_of<ID3D12CommandQueue>(), commandQueue.put_void()));
+  // create a winml learning device based on that dx12 queue
+  auto factory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
+  com_ptr<::IUnknown> spUnk;
+  WINML_EXPECT_HRESULT_SUCCEEDED(factory->CreateFromD3D12CommandQueue(commandQueue.get(), spUnk.put()));
+  auto learningDevice = spUnk.as<LearningModelDevice>();
+  // create a winml session from that dx device
+  LearningModelSession session(model, learningDevice);
+  // now lets try and do some XAML/d2d on that same device, first prealloc a VideoFrame
+  VideoFrame frame = VideoFrame::CreateAsDirect3D11SurfaceBacked(
+      DirectXPixelFormat::B8G8R8A8UIntNormalized,
+      224,
+      224,
+      session.Device().Direct3D11Device());
+  // create a D2D factory
+  D2D1_FACTORY_OPTIONS options = {};
+  com_ptr<ID2D1Factory> d2dFactory;
+  WINML_EXPECT_HRESULT_SUCCEEDED(D2D1CreateFactory(D2D1_FACTORY_TYPE_SINGLE_THREADED, __uuidof(ID2D1Factory), &options, d2dFactory.put_void()));
+  // grab the dxgi surface back from our video frame
+  com_ptr<IDXGISurface> dxgiSurface;
+  com_ptr<IDirect3DDxgiInterfaceAccess> dxgiInterfaceAccess = frame.Direct3DSurface().as<IDirect3DDxgiInterfaceAccess>();
+  WINML_EXPECT_HRESULT_SUCCEEDED(dxgiInterfaceAccess->GetInterface(__uuidof(IDXGISurface), dxgiSurface.put_void()));
+  // and try and use our surface to create a render targer
+  com_ptr<ID2D1RenderTarget> renderTarget;
+  D2D1_RENDER_TARGET_PROPERTIES props = D2D1::RenderTargetProperties();
+  props.pixelFormat = D2D1::PixelFormat(
+      DXGI_FORMAT_B8G8R8A8_UNORM,
+      D2D1_ALPHA_MODE_IGNORE);
+  WINML_EXPECT_HRESULT_SUCCEEDED(d2dFactory->CreateDxgiSurfaceRenderTarget(
+      dxgiSurface.get(),
+      props,
+      renderTarget.put()));
 }
 
-TEST_F(ScenarioCppWinrtGpuSkipEdgeCoreTest, D2DInterop)
-{
-    // load a model (model.onnx == squeezenet[1,3,224,224])
-    std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
-    LearningModel model = LearningModel::LoadFromFilePath(filePath);
-    // create a dx12 device
-    com_ptr<ID3D12Device1> device = nullptr;
-    EXPECT_HRESULT_SUCCEEDED(D3D12CreateDevice(NULL, D3D_FEATURE_LEVEL_11_0, __uuidof(ID3D12Device1), device.put_void()));
-    // now create a command queue from it
-    com_ptr<ID3D12CommandQueue> commandQueue = nullptr;
-    D3D12_COMMAND_QUEUE_DESC queueDesc = {};
-    queueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT;
-    EXPECT_HRESULT_SUCCEEDED(device->CreateCommandQueue(&queueDesc, winrt::guid_of<ID3D12CommandQueue>(), commandQueue.put_void()));
-    // create a winml learning device based on that dx12 queue
-    auto factory = get_activation_factory<LearningModelDevice, ILearningModelDeviceFactoryNative>();
-    com_ptr<::IUnknown> spUnk;
-    EXPECT_HRESULT_SUCCEEDED(factory->CreateFromD3D12CommandQueue(commandQueue.get(), spUnk.put()));
-    auto learningDevice = spUnk.as<LearningModelDevice>();
-    // create a winml session from that dx device
-    LearningModelSession session(model, learningDevice);
-    // now lets try and do some XAML/d2d on that same device, first prealloc a VideoFrame
-    VideoFrame frame = VideoFrame::CreateAsDirect3D11SurfaceBacked(
-        DirectXPixelFormat::B8G8R8A8UIntNormalized,
-        224,
-        224,
-        session.Device().Direct3D11Device()
-        );
-    // create a D2D factory
-    D2D1_FACTORY_OPTIONS options = {};
-    com_ptr<ID2D1Factory> d2dFactory;
-    EXPECT_HRESULT_SUCCEEDED(D2D1CreateFactory(D2D1_FACTORY_TYPE_SINGLE_THREADED, __uuidof(ID2D1Factory), &options, d2dFactory.put_void()));
-    // grab the dxgi surface back from our video frame
-    com_ptr<IDXGISurface> dxgiSurface;
-    com_ptr<IDirect3DDxgiInterfaceAccess> dxgiInterfaceAccess = frame.Direct3DSurface().as<IDirect3DDxgiInterfaceAccess>();
-    EXPECT_HRESULT_SUCCEEDED(dxgiInterfaceAccess->GetInterface(__uuidof(IDXGISurface), dxgiSurface.put_void()));
-    // and try and use our surface to create a render targer
-    com_ptr<ID2D1RenderTarget> renderTarget;
-    D2D1_RENDER_TARGET_PROPERTIES props = D2D1::RenderTargetProperties();
-    props.pixelFormat = D2D1::PixelFormat(
-        DXGI_FORMAT_B8G8R8A8_UNORM,
-        D2D1_ALPHA_MODE_IGNORE
-    );
-    EXPECT_HRESULT_SUCCEEDED(d2dFactory->CreateDxgiSurfaceRenderTarget(
-        dxgiSurface.get(),
-        props,
-        renderTarget.put()));
-}
+const ScenarioTestApi& getapi() {
+  static constexpr ScenarioTestApi api =
+      {
+        ScenarioCppWinrtTestSetup,
+        ScenarioCppWinrtGpuTestSetup,
+        ScenarioCppWinrtGpuSkipEdgeCoreTestSetup,
+        Sample1,
+        Scenario1LoadBindEvalDefault,
+        Scenario2LoadModelFromStream,
+        Scenario5AsyncEval,
+        Scenario7EvalWithNoBind,
+        Scenario8SetDeviceSampleDefault,
+        Scenario8SetDeviceSampleCPU,
+        Scenario17DevDiagnostics,
+        Scenario22ImageBindingAsCPUTensor,
+        QuantizedModels,
+        EncryptedStream,
+        Scenario3SoftwareBitmapInputBinding,
+        Scenario6BindWithProperties,
+        Scenario8SetDeviceSampleDefaultDirectX,
+        Scenario8SetDeviceSampleMinPower,
+        Scenario8SetDeviceSampleMaxPerf,
+        Scenario8SetDeviceSampleMyCameraDevice,
+        Scenario8SetDeviceSampleCustomCommandQueue,
+        Scenario9LoadBindEvalInputTensorGPU,
+        Scenario13SingleModelOnCPUandGPU,
+        Scenario11FreeDimensionsTensor,
+        Scenario11FreeDimensionsImage,
+        Scenario14RunModelSwapchain,
+        Scenario20aLoadBindEvalCustomOperatorCPU,
+        Scenario20bLoadBindEvalReplacementCustomOperatorCPU,
+        Scenario21RunModel2ChainZ,
+        Scenario22ImageBindingAsGPUTensor,
+        MsftQuantizedModels,
+        SyncVsAsync,
+        CustomCommandQueueWithFence,
+        ReuseVideoFrame,
+        DeviceLostRecovery,
+        Scenario8SetDeviceSampleD3D11Device,
+        D2DInterop,
+      };
+  return api;
+}
\ No newline at end of file
diff --git a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h
new file mode 100644
index 0000000000000..9e49cf1d3752d
--- /dev/null
+++ b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h
@@ -0,0 +1,85 @@
+#include "test.h"
+struct ScenarioTestApi
+{
+    SetupTest ScenarioCppWinrtTestSetup;
+    SetupTest ScenarioCppWinrtGpuTestSetup;
+    SetupTest ScenarioCppWinrtGpuSkipEdgeCoreTestSetup;
+    VoidTest Sample1;
+    VoidTest Scenario1LoadBindEvalDefault;
+    VoidTest Scenario2LoadModelFromStream;
+    VoidTest Scenario5AsyncEval;
+    VoidTest Scenario7EvalWithNoBind;
+    VoidTest Scenario8SetDeviceSampleDefault;
+    VoidTest Scenario8SetDeviceSampleCPU;
+    VoidTest Scenario17DevDiagnostics;
+    VoidTest DISABLED_Scenario22ImageBindingAsCPUTensor;
+    VoidTest QuantizedModels;
+    VoidTest EncryptedStream;
+    VoidTest Scenario3SoftwareBitmapInputBinding;
+    VoidTest Scenario6BindWithProperties;
+    VoidTest Scenario8SetDeviceSampleDefaultDirectX;
+    VoidTest Scenario8SetDeviceSampleMinPower;
+    VoidTest Scenario8SetDeviceSampleMaxPerf;
+    VoidTest Scenario8SetDeviceSampleMyCameraDevice;
+    VoidTest Scenario8SetDeviceSampleCustomCommandQueue;
+    VoidTest DISABLED_Scenario9LoadBindEvalInputTensorGPU;
+    VoidTest Scenario13SingleModelOnCPUandGPU;
+    VoidTest Scenario11FreeDimensionsTensor;
+    VoidTest Scenario11FreeDimensionsImage;
+    VoidTest Scenario14RunModelSwapchain;
+    VoidTest Scenario20aLoadBindEvalCustomOperatorCPU;
+    VoidTest Scenario20bLoadBindEvalReplacementCustomOperatorCPU;
+    VoidTest DISABLED_Scenario21RunModel2ChainZ;
+    VoidTest DISABLED_Scenario22ImageBindingAsGPUTensor;
+    VoidTest MsftQuantizedModels;
+    VoidTest DISABLED_SyncVsAsync;
+    VoidTest DISABLED_CustomCommandQueueWithFence;
+    VoidTest DISABLED_ReuseVideoFrame;
+    VoidTest DeviceLostRecovery;
+    VoidTest Scenario8SetDeviceSampleD3D11Device;
+    VoidTest D2DInterop;
+};
+const ScenarioTestApi& getapi();
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(ScenarioCppWinrtTest, ScenarioCppWinrtTestSetup)
+WINML_TEST(ScenarioCppWinrtTest, Sample1)
+WINML_TEST(ScenarioCppWinrtTest, Scenario1LoadBindEvalDefault)
+WINML_TEST(ScenarioCppWinrtTest, Scenario2LoadModelFromStream)
+WINML_TEST(ScenarioCppWinrtTest, Scenario5AsyncEval)
+WINML_TEST(ScenarioCppWinrtTest, Scenario7EvalWithNoBind)
+WINML_TEST(ScenarioCppWinrtTest, Scenario8SetDeviceSampleDefault)
+WINML_TEST(ScenarioCppWinrtTest, Scenario8SetDeviceSampleCPU)
+WINML_TEST(ScenarioCppWinrtTest, Scenario17DevDiagnostics)
+WINML_TEST(ScenarioCppWinrtTest, DISABLED_Scenario22ImageBindingAsCPUTensor)
+WINML_TEST(ScenarioCppWinrtTest, QuantizedModels)
+WINML_TEST(ScenarioCppWinrtTest, EncryptedStream)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(ScenarioCppWinrtGpuTest, ScenarioCppWinrtGpuTestSetup)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario3SoftwareBitmapInputBinding)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario6BindWithProperties)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleDefaultDirectX)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMinPower)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMaxPerf)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMyCameraDevice)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleCustomCommandQueue)
+WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_Scenario9LoadBindEvalInputTensorGPU)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario13SingleModelOnCPUandGPU)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario11FreeDimensionsTensor)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario11FreeDimensionsImage)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario14RunModelSwapchain)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario20aLoadBindEvalCustomOperatorCPU)
+WINML_TEST(ScenarioCppWinrtGpuTest, Scenario20bLoadBindEvalReplacementCustomOperatorCPU)
+WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_Scenario21RunModel2ChainZ)
+WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_Scenario22ImageBindingAsGPUTensor)
+WINML_TEST(ScenarioCppWinrtGpuTest, MsftQuantizedModels)
+WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_SyncVsAsync)
+WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_CustomCommandQueueWithFence)
+WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_ReuseVideoFrame)
+WINML_TEST(ScenarioCppWinrtGpuTest, DeviceLostRecovery)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(ScenarioCppWinrtGpuSkipEdgeCoreTest, ScenarioCppWinrtGpuSkipEdgeCoreTestSetup)
+WINML_TEST(ScenarioCppWinrtGpuSkipEdgeCoreTest, Scenario8SetDeviceSampleD3D11Device )
+WINML_TEST(ScenarioCppWinrtGpuSkipEdgeCoreTest, D2DInterop)
+WINML_TEST_CLASS_END()
\ No newline at end of file

From cc75e5a16243f031d6160d576417ae1999a4ab87 Mon Sep 17 00:00:00 2001
From: Ashwini Khade <askhade@microsoft.com>
Date: Mon, 13 Jan 2020 10:52:46 -0800
Subject: [PATCH 47/75] update quantization doc (#2783)

* update documentation for quantization script

* plus some spell corrections
---
 .../python/tools/quantization/README.md       | 27 ++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/onnxruntime/python/tools/quantization/README.md b/onnxruntime/python/tools/quantization/README.md
index abbd30fdd8605..37d992c0b2a3e 100644
--- a/onnxruntime/python/tools/quantization/README.md
+++ b/onnxruntime/python/tools/quantization/README.md
@@ -1,5 +1,30 @@
 # Quantization tool Overview
-This tool supports quantization of an onnx model. quantize() takes a model in ModelProto format and returns the quantized model in ModelProto format.
+This tool supports 8 bit linear quantization of an onnx model. quantize() takes a model in ModelProto format and returns the quantized model in ModelProto format.
+Today ORT does not guarantee support for E2E model quantization, meaning all ONNX ops do not have support for 8 bit data types therefore only the supported ops in the model are quantized. For rest of the ops inputs are reconverted to FP32.
+
+List of Supported Quantized Ops:
+The following ops were chosen as phase 1 ops because in most of the CNN models these ops consume most amount of compute and power and therefore there is benefit in quantizing these ops to get perf benefits.
+ * Convolution
+ * Matmul
+ * Data type agnostic ops like transpose, identity etc ( Note: special quantization is not done for these ops. )
+
+ ## Quantization specifics
+ ONNX implements 8 bit linear quantization. During quantization the floating point real values are mapped to a 8 bit quantization space and it is of the form :
+ VAL_fp32 = Scale * (VAL_quantized - Zero_point)
+ 
+ Scale is a positive real number used to map the floating point numbers to a quantization space. It is calculated as follows : 
+ For unsigned 8 bit
+ ```
+ scale = (data_rage_max - data_range_min) / (quantization_range_max - quantization_range_min)
+ ```
+
+ For signed 8 bit
+ ```
+ scale = Abs(data_rage_max, data_range_min) * 2 / (quantization_range_max - quantization_range_min)
+ ```
+
+ Zero point represents zero in quantization space. It is important that floating point zero value be exactly representable in quantization space. This is because in lot of CNNs, zero padding is used and if after quantization it is not possible to represent 0 uniquely then it will lead to accuracy errors.
+
 
 ## Quantize an ONNX model
 ```python

From 6ad3e74f4afa127c3179d5be466284e8771b3ea4 Mon Sep 17 00:00:00 2001
From: Ori Levari <ori.levari@microsoft.com>
Date: Mon, 13 Jan 2020 13:56:14 -0800
Subject: [PATCH 48/75] Filter CPU case for IsFloat16Supported (#2802)

---
 winml/lib/Api.Image/DeviceHelpers.cpp | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/winml/lib/Api.Image/DeviceHelpers.cpp b/winml/lib/Api.Image/DeviceHelpers.cpp
index 67824624a7b0a..3174f61fad619 100644
--- a/winml/lib/Api.Image/DeviceHelpers.cpp
+++ b/winml/lib/Api.Image/DeviceHelpers.cpp
@@ -9,6 +9,7 @@
 #include <d3d11on12.h>
 #include <wil/winrt.h>
 #include "inc/DeviceHelpers.h"
+#include "LearningModelDevice.h"
 
 namespace DeviceHelpers {
 constexpr uint32_t c_intelVendorId = 0x8086;
@@ -133,6 +134,10 @@ static HRESULT IsFloat16Blocked(ID3D12Device& device, bool* isBlocked) {
 }
 
 bool IsFloat16Supported(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device) {
+  auto modelImpl = device.as<winmlp::LearningModelDevice>();
+  if (modelImpl->IsCpuDevice()) {
+    return true;
+  }
   winrt::com_ptr<ID3D12Device> d3d12Device;
   if (FAILED(GetD3D12Device(device, d3d12Device.put()))) {
     return false;

From 7c6242b0244731433b04d37d151271b322d3cb53 Mon Sep 17 00:00:00 2001
From: Ashwini Khade <askhade@microsoft.com>
Date: Mon, 13 Jan 2020 14:05:38 -0800
Subject: [PATCH 49/75] update default optimization level + fix gemm_activation
 fusion (#2791)

* update defualt optimization level + fix gemm_activation fusion

* fix typo

* add unit test and incorporate review comments

* fix test comment
---
 .../SessionOptions.cs                         |   4 +--
 .../InferenceTest.cs                          |   2 +-
 .../Program.cs                                |   4 +--
 docs/ONNX_Runtime_Graph_Optimizations.md      |   4 ++-
 docs/ONNX_Runtime_Perf_Tuning.md              |   4 +--
 onnxruntime/core/framework/session_options.h  |   2 +-
 .../core/optimizer/gemm_activation_fusion.cc  |   4 ++-
 .../python/onnxruntime_pybind_state.cc        |   6 ++---
 onnxruntime/test/onnx/main.cc                 |   2 +-
 .../test/optimizer/graph_transform_test.cc    |  17 +++++++++++++
 .../test/perftest/test_configuration.h        |   2 +-
 .../test/python/onnxruntime_test_python.py    |   8 +++---
 .../gemm_activation_fusion.onnx               |  24 ++++++++++++++++++
 .../test_data_set_0/input_0.pb                |   1 +
 .../test_data_set_0/input_1.pb                |   1 +
 .../test_data_set_0/input_2.pb                | Bin 0 -> 27 bytes
 .../test_data_set_0/output_0.pb               |   1 +
 17 files changed, 67 insertions(+), 19 deletions(-)
 create mode 100644 onnxruntime/test/testdata/transform/gemm_activation_fusion/gemm_activation_fusion.onnx
 create mode 100644 onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_0.pb
 create mode 100644 onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_1.pb
 create mode 100644 onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_2.pb
 create mode 100644 onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/output_0.pb

diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
index c2a5920741d4d..6f24f49634da0 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/SessionOptions.cs
@@ -345,7 +345,7 @@ public int InterOpNumThreads
         private int _interOpNumThreads = 0; // set to what is set in C++ SessionOptions by default;
 
         /// <summary>
-        /// Sets the graph optimization level for the session. Default is set to ORT_ENABLE_BASIC.        
+        /// Sets the graph optimization level for the session. Default is set to ORT_ENABLE_ALL.
         /// </summary>
         public GraphOptimizationLevel GraphOptimizationLevel
         {
@@ -359,7 +359,7 @@ public GraphOptimizationLevel GraphOptimizationLevel
                 _graphOptimizationLevel = value;
             }
         }
-        private GraphOptimizationLevel _graphOptimizationLevel = GraphOptimizationLevel.ORT_ENABLE_BASIC;
+        private GraphOptimizationLevel _graphOptimizationLevel = GraphOptimizationLevel.ORT_ENABLE_ALL;
 
         /// <summary>
         /// Sets the execution mode for the session. Default is set to ORT_SEQUENTIAL.
diff --git a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
index e60cbb8aacf7f..7b221f5e33673 100644
--- a/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
+++ b/csharp/test/Microsoft.ML.OnnxRuntime.Tests/InferenceTest.cs
@@ -41,7 +41,7 @@ public void TestSessionOptions()
                 Assert.Equal(LogLevel.Verbose, opt.LogVerbosityLevel);
                 Assert.Equal(0, opt.IntraOpNumThreads);
                 Assert.Equal(0, opt.InterOpNumThreads);
-                Assert.Equal(GraphOptimizationLevel.ORT_ENABLE_BASIC, opt.GraphOptimizationLevel);
+                Assert.Equal(GraphOptimizationLevel.ORT_ENABLE_ALL, opt.GraphOptimizationLevel);
 
                 // try setting options 
                 opt.ExecutionMode = ExecutionMode.ORT_PARALLEL;
diff --git a/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs b/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
index de589114e7bad..a9af0449bd15c 100644
--- a/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
+++ b/csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs
@@ -32,8 +32,8 @@ class CommandOptions
         [Option('p', Required = false, HelpText = "Run with parallel exection. Default is false")]
         public bool ParallelExecution { get; set; } = false;
 
-        [Option('o', "optimization_level", Required = false, HelpText = "Optimization Level. Default is 1, partial optimization.")]
-        public GraphOptimizationLevel OptimizationLevel { get; set; } = GraphOptimizationLevel.ORT_ENABLE_BASIC;
+        [Option('o', "optimization_level", Required = false, HelpText = "Optimization Level. Default is 99, all optimization.")]
+        public GraphOptimizationLevel OptimizationLevel { get; set; } = GraphOptimizationLevel.ORT_ENABLE_ALL;
     }
 
     class Program
diff --git a/docs/ONNX_Runtime_Graph_Optimizations.md b/docs/ONNX_Runtime_Graph_Optimizations.md
index 0c347f87e477d..2d124420db980 100644
--- a/docs/ONNX_Runtime_Graph_Optimizations.md
+++ b/docs/ONNX_Runtime_Graph_Optimizations.md
@@ -15,9 +15,11 @@ Graph optimizations are divided in three levels:
 
 The optimizations belonging to one level are performed after the optimizations of the previous level have been applied (e.g., extended optimizations are applied after basic optimizations have been applied).
 
+**All optimizations are enabled by default.**
+
 ### Basic Graph Optimizations
 
-These are semantics-preserving graph rewrites which remove redundant nodes and redundant computation. These optimizations are enabled by default. They run before graph partitioning and thus apply to all the execution providers. Available basic graph optimizations are as follows:
+These are semantics-preserving graph rewrites which remove redundant nodes and redundant computation. They run before graph partitioning and thus apply to all the execution providers. Available basic graph optimizations are as follows:
 
 * Constant Folding: Statically computes parts of the graph that rely only on constant initializers. This eliminates the need to compute them during runtime.
 
diff --git a/docs/ONNX_Runtime_Perf_Tuning.md b/docs/ONNX_Runtime_Perf_Tuning.md
index 2788ae041b85e..aaab8f8422fe5 100644
--- a/docs/ONNX_Runtime_Perf_Tuning.md
+++ b/docs/ONNX_Runtime_Perf_Tuning.md
@@ -86,9 +86,9 @@ sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL
 * Sequential vs Parallel Execution
   * `sess_options.execution_mode = rt.ExecutionMode.ORT_SEQUENTIAL` controls whether then operators in the graph should run sequentially or in parallel. Usually when a model has many branches, setting this option to false will provide better performance.
   * When `sess_options.execution_mode = rt.ExecutionMode.ORT_PARALLEL`, you can set `sess_options.inter_op_num_threads` to control the
-
 number of threads used to parallelize the execution of the graph (across nodes).
-* sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL. Default is ORT_ENABLE_BASIC(1). Please see [onnxruntime_c_api.h](../include/onnxruntime/core/session/onnxruntime_c_api.h#L241)  (enum GraphOptimizationLevel) for the full list of all optimization levels. For details regarding available optimizations and usage please refer to the [Graph Optimizations Doc](../docs/ONNX_Runtime_Graph_Optimizations.md).
+
+* sess_options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL. Default is already ORT_ENABLE_ALL(99). Please see [onnxruntime_c_api.h](../include/onnxruntime/core/session/onnxruntime_c_api.h#L241)  (enum GraphOptimizationLevel) for the full list of all optimization levels. For details regarding available optimizations and usage please refer to the [Graph Optimizations Doc](../docs/ONNX_Runtime_Graph_Optimizations.md).
 
 ### MKL_DNN/nGraph/MKL_ML Execution Provider
 MKL_DNN, MKL_ML and nGraph all depends on openmp for parallization. For those execution providers, we need to use the openmp enviroment variable to tune the performance.
diff --git a/onnxruntime/core/framework/session_options.h b/onnxruntime/core/framework/session_options.h
index aaefc50cc6c9c..1c02f9f0ebbc2 100644
--- a/onnxruntime/core/framework/session_options.h
+++ b/onnxruntime/core/framework/session_options.h
@@ -52,7 +52,7 @@ struct SessionOptions {
   unsigned max_num_graph_transformation_steps = 10;  // TODO choose a good default here?
 
   // set graph optimization level
-  TransformerLevel graph_optimization_level = TransformerLevel::Level1;
+  TransformerLevel graph_optimization_level = TransformerLevel::Level3;
 
   // controls the size of the thread pool used to parallelize the execution of tasks within individual nodes (ops)
   int intra_op_num_threads = 0;
diff --git a/onnxruntime/core/optimizer/gemm_activation_fusion.cc b/onnxruntime/core/optimizer/gemm_activation_fusion.cc
index 05d1c864b770b..c6c5e5316c28a 100644
--- a/onnxruntime/core/optimizer/gemm_activation_fusion.cc
+++ b/onnxruntime/core/optimizer/gemm_activation_fusion.cc
@@ -68,7 +68,9 @@ Status GemmActivationFusion::ApplyImpl(Graph& graph, bool& modified, int graph_l
     if (act_node.OpType() == "LeakyRelu") {
       const NodeAttributes& attrs = act_node.GetAttributes();
       for (const auto& attr : attrs) {
-        fused_gemm.AddAttribute("leaky_relu_" + attr.first, attr.second);
+        AttributeProto fused_gemm_attr(attr.second);
+        fused_gemm_attr.set_name("leaky_relu_" + attr.first);
+        fused_gemm.AddAttribute("leaky_relu_" + attr.first, fused_gemm_attr);
       }
     }
 
diff --git a/onnxruntime/python/onnxruntime_pybind_state.cc b/onnxruntime/python/onnxruntime_pybind_state.cc
index e48c03323cc55..ba6035ba2da70 100644
--- a/onnxruntime/python/onnxruntime_pybind_state.cc
+++ b/onnxruntime/python/onnxruntime_pybind_state.cc
@@ -587,7 +587,7 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
       .def_property(
           "graph_optimization_level",
           [](const SessionOptions* options) -> GraphOptimizationLevel {
-            GraphOptimizationLevel retval = ORT_ENABLE_BASIC;
+            GraphOptimizationLevel retval = ORT_ENABLE_ALL;
             switch (options->graph_optimization_level) {
               case onnxruntime::TransformerLevel::Default:
                 retval = ORT_DISABLE_ALL;
@@ -602,8 +602,8 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
                 retval = ORT_ENABLE_ALL;
                 break;
               default:
-                retval = ORT_ENABLE_BASIC;
-                LOGS_DEFAULT(WARNING) << "Got invalid graph optimization level; defaulting to ORT_ENABLE_BASIC";
+                retval = ORT_ENABLE_ALL;
+                LOGS_DEFAULT(WARNING) << "Got invalid graph optimization level; defaulting to ORT_ENABLE_ALL";
                 break;
             }
             return retval;
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index ff881510714b4..6113657cae7ee 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -41,7 +41,7 @@ void usage() {
       "Default: 'cpu'.\n"
       "\t-x: Use parallel executor, default (without -x): sequential executor.\n"
       "\t-d [device_id]: Specifies the device id for multi-device (e.g. GPU). The value should > 0\n"
-      "\t-o [optimization level]: Default is 1. Valid values are 0 (disable), 1 (basic), 2 (extended), 99 (all).\n"
+      "\t-o [optimization level]: Default is 99. Valid values are 0 (disable), 1 (basic), 2 (extended), 99 (all).\n"
       "\t\tPlease see onnxruntime_c_api.h (enum GraphOptimizationLevel) for the full list of all optimization levels. "
       "\n"
       "\t-h: help\n"
diff --git a/onnxruntime/test/optimizer/graph_transform_test.cc b/onnxruntime/test/optimizer/graph_transform_test.cc
index bbfc1fcc0dd68..46e3682c0c763 100644
--- a/onnxruntime/test/optimizer/graph_transform_test.cc
+++ b/onnxruntime/test/optimizer/graph_transform_test.cc
@@ -609,6 +609,23 @@ TEST(GraphTransformationTests, Gemm_Relu_three_input) {
   std::map<std::string, int> op_to_count = CountOpsInGraph(graph);
   ASSERT_TRUE(op_to_count["Relu"] == 0);
 }
+
+TEST(GraphTransformationTests, Gemm_LeakyRelu_Fusion) {
+  auto model_uri = MODEL_FOLDER "gemm_activation_fusion/gemm_activation_fusion.onnx";
+
+  std::shared_ptr<Model> p_model;
+  ASSERT_TRUE(Model::Load(model_uri, p_model, nullptr, DefaultLoggingManager().DefaultLogger()).IsOK());
+  Graph& graph = p_model->MainGraph();
+  std::map<std::string, int> op_to_count1 = CountOpsInGraph(graph);
+  onnxruntime::GraphTransformerManager graph_transformation_mgr{5};
+  graph_transformation_mgr.Register(onnxruntime::make_unique<GemmActivationFusion>(), TransformerLevel::Level2);
+  ASSERT_TRUE(graph_transformation_mgr.ApplyTransformers(graph, TransformerLevel::Level2, DefaultLoggingManager().DefaultLogger()).IsOK());
+
+  std::map<std::string, int> op_to_count = CountOpsInGraph(graph);
+  ASSERT_TRUE(op_to_count["LeakyRelu"] == 0);
+  ASSERT_TRUE(op_to_count["Gemm"] == 0);
+  ASSERT_TRUE(op_to_count["FusedGemm"] == 1);
+}
 #endif
 
 TEST(GraphTransformationTests, FuseConvBnAddMulFloat16) {
diff --git a/onnxruntime/test/perftest/test_configuration.h b/onnxruntime/test/perftest/test_configuration.h
index 9d0bf46774057..36186e129f0f8 100644
--- a/onnxruntime/test/perftest/test_configuration.h
+++ b/onnxruntime/test/perftest/test_configuration.h
@@ -47,7 +47,7 @@ struct RunConfig {
   ExecutionMode execution_mode{ExecutionMode::ORT_SEQUENTIAL};
   int intra_op_num_threads{0};
   int inter_op_num_threads{0};
-  GraphOptimizationLevel optimization_level{ORT_ENABLE_EXTENDED};
+  GraphOptimizationLevel optimization_level{ORT_ENABLE_ALL};
   std::basic_string<ORTCHAR_T> optimized_model_path;
 };
 
diff --git a/onnxruntime/test/python/onnxruntime_test_python.py b/onnxruntime/test/python/onnxruntime_test_python.py
index 0a0060767ebea..21b0fc8a8c457 100644
--- a/onnxruntime/test/python/onnxruntime_test_python.py
+++ b/onnxruntime/test/python/onnxruntime_test_python.py
@@ -521,10 +521,10 @@ def test_run_model_mlnet(self):
 
     def testGraphOptimizationLevel(self):
         opt = onnxrt.SessionOptions()
-        self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_BASIC)
-            # default should be basic optimization
-        opt.graph_optimization_level = onnxrt.GraphOptimizationLevel.ORT_ENABLE_ALL
+        # default should be all optimizations optimization
         self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_ALL)
+        opt.graph_optimization_level = onnxrt.GraphOptimizationLevel.ORT_ENABLE_EXTENDED
+        self.assertEqual(opt.graph_optimization_level, onnxrt.GraphOptimizationLevel.ORT_ENABLE_EXTENDED)
         sess = onnxrt.InferenceSession(self.get_name("logicaland.onnx"), sess_options=opt)
         a = np.array([[True, True], [False, False]], dtype=np.bool)
         b = np.array([[True, False], [True, False]], dtype=np.bool)
@@ -622,7 +622,7 @@ def testLoadingSessionOptionsFromModel(self):
 
         finally:
             # Make sure the usage of the feature is disabled after this test
-            os.environ['ORT_LOAD_CONFIG_FROM_MODEL'] = str(0)       
+            os.environ['ORT_LOAD_CONFIG_FROM_MODEL'] = str(0)
         
 if __name__ == '__main__':
     unittest.main()
diff --git a/onnxruntime/test/testdata/transform/gemm_activation_fusion/gemm_activation_fusion.onnx b/onnxruntime/test/testdata/transform/gemm_activation_fusion/gemm_activation_fusion.onnx
new file mode 100644
index 0000000000000..87dd469343005
--- /dev/null
+++ b/onnxruntime/test/testdata/transform/gemm_activation_fusion/gemm_activation_fusion.onnx
@@ -0,0 +1,24 @@
+onnx-helper:“

+
+
a
+
b
+
c
y"Gemm
+
+
y
z"	LeakyRelugemm_activation_fusionZ
+
a
+

+
+Z
+
b
+

+
+Z
+
c
+

+

+b
+
z
+

+
+B
+ai.onnx	
\ No newline at end of file
diff --git a/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_0.pb b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_0.pb
new file mode 100644
index 0000000000000..ee690bab01f44
--- /dev/null
+++ b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_0.pb
@@ -0,0 +1 @@
+
B
aJ<?¦7?³N?w}?HéØ>QY%?nà><Kd?š²v?rRÄ>~®J?¨e?^k?çól?Z{‘=
\ No newline at end of file
diff --git a/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_1.pb b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_1.pb
new file mode 100644
index 0000000000000..6945e22c76d0a
--- /dev/null
+++ b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_1.pb
@@ -0,0 +1 @@
+
B
bJPÙp²=
¡¥<“&U?H5G?¹^?»†z?¨•L?Gì>ÃÐG?Ý9ò=ÝÑ#?4Ë>ÒÕq?Ú—?’NÔ>Õs‡>.4F?‰Œé>ã„?í™<
\ No newline at end of file
diff --git a/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_2.pb b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/input_2.pb
new file mode 100644
index 0000000000000000000000000000000000000000..7077e4ffe336e33efce98d917289ebaaea95ee96
GIT binary patch
literal 27
Ucmd;J<X{nCbYe{Q5@0|B02bB(CjbBd

literal 0
HcmV?d00001

diff --git a/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/output_0.pb b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/output_0.pb
new file mode 100644
index 0000000000000..f53ed3aaf25e2
--- /dev/null
+++ b/onnxruntime/test/testdata/transform/gemm_activation_fusion/test_data_set_0/output_0.pb
@@ -0,0 +1 @@
+
B
zJ0cÛý?ÎC¡?%Äð?ƒá~?´Æ@D¾œ?\Ø@�Œ?ípó?½ßŽ?
Nï?Â	˜?
\ No newline at end of file

From 07502ec14eab0ab17807fd23be7c028be15dc519 Mon Sep 17 00:00:00 2001
From: Yingge WAN <y-wan@users.noreply.github.com>
Date: Tue, 14 Jan 2020 06:37:11 +0800
Subject: [PATCH 50/75] Fix dnnl wheel package name (#2823)

* Append '-dnnl' to whl package name when --use_dnnl

* Update build.py
---
 setup.py                | 4 ++++
 tools/ci_build/build.py | 6 ++++--
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/setup.py b/setup.py
index e0dfb4e729883..4af75ccc873da 100644
--- a/setup.py
+++ b/setup.py
@@ -34,6 +34,10 @@
 elif '--use_ngraph' in sys.argv:
     package_name = 'onnxruntime-ngraph'
     sys.argv.remove('--use_ngraph')
+    
+elif '--use_dnnl' in sys.argv:
+    package_name = 'onnxruntime-dnnl'
+    sys.argv.remove('--use_dnnl')
 
 elif '--use_openvino' in sys.argv:
     package_name = 'onnxruntime-openvino'
diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index c7aad8e0e2760..32a810ed7af6f 100755
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -705,7 +705,7 @@ def nuphar_run_python_tests(build_dir, configs):
         run_subprocess([sys.executable, 'onnxruntime_test_python_nuphar.py'], cwd=cwd, dll_path=dll_path)
 
 
-def build_python_wheel(source_dir, build_dir, configs, use_cuda, use_ngraph, use_tensorrt, use_openvino, use_nuphar, nightly_build = False):
+def build_python_wheel(source_dir, build_dir, configs, use_cuda, use_ngraph, use_dnnl, use_tensorrt, use_openvino, use_nuphar, nightly_build = False):
     for config in configs:
         cwd = get_config_build_dir(build_dir, config)
         if is_windows():
@@ -719,6 +719,8 @@ def build_python_wheel(source_dir, build_dir, configs, use_cuda, use_ngraph, use
             args.append('--use_cuda')
         elif use_ngraph:
             args.append('--use_ngraph')
+        elif use_dnnl:
+            args.append('--use_dnnl')
         elif use_openvino:
             args.append('--use_openvino')
         elif use_nuphar:
@@ -969,7 +971,7 @@ def main():
     if args.build:
         if args.build_wheel:
             nightly_build = bool(os.getenv('NIGHTLY_BUILD') == '1')
-            build_python_wheel(source_dir, build_dir, configs, args.use_cuda, args.use_ngraph, args.use_tensorrt, args.use_openvino, args.use_nuphar, nightly_build)
+            build_python_wheel(source_dir, build_dir, configs, args.use_cuda, args.use_ngraph, args.use_dnnl, args.use_tensorrt, args.use_openvino, args.use_nuphar, nightly_build)
 
     if args.gen_doc and (args.build or args.test):
         generate_documentation(source_dir, build_dir, configs)

From 723cf8379300c5d054635135043d52c006d3883e Mon Sep 17 00:00:00 2001
From: Qing <cwq1913@gmail.com>
Date: Tue, 14 Jan 2020 06:37:32 +0800
Subject: [PATCH 51/75] Update Ubuntu & TensorRT version  in README (#2820)

Dockerfile.tensorrt is using nvcr.io/nvidia/tensorrt:19.09-py3 as base Image, update Ubuntu and TensorRT version according to
https://docs.nvidia.com/deeplearning/sdk/tensorrt-container-release-notes/rel_19-09.html#rel_19-09
---
 dockerfiles/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dockerfiles/README.md b/dockerfiles/README.md
index d99aecff57020..ae32c1fcac6cc 100644
--- a/dockerfiles/README.md
+++ b/dockerfiles/README.md
@@ -80,7 +80,7 @@ Use `docker pull` with any of the images and tags below to pull an image and try
   ```
 
 ## TensorRT
-**Ubuntu 16.04, TensorRT 5.0.2**
+**Ubuntu 18.04, CUDA 10.1.243, TensorRT 6.0.1**
 
 1. Build the docker image from the Dockerfile in this repository.
   ```

From 3e8768d1b8f26d01b7d375d8872c3b3f0792f143 Mon Sep 17 00:00:00 2001
From: Jeff Bloomfield <jeffbloo@ntdev.microsoft.com>
Date: Mon, 13 Jan 2020 17:58:02 -0800
Subject: [PATCH 52/75] Merge fixes

---
 .../dml/DmlExecutionProvider/src/ExecutionProvider.h          | 2 +-
 .../providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h | 2 +-
 .../providers/dml/DmlExecutionProvider/src/GraphPartitioner.h | 4 ++--
 .../dml/OperatorAuthorHelper/SchemaInferenceOverrider.h       | 2 +-
 winml/adapter/AbiCustomRegistryImpl.cpp                       | 4 +++-
 winml/adapter/AbiCustomRegistryImpl.h                         | 1 +
 6 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
index 52986d1faa194..4c14420441a0b 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/ExecutionProvider.h
@@ -170,7 +170,7 @@ namespace Dml
         std::shared_ptr<CPUAllocator> m_cpuInputAllocator;
         std::shared_ptr<CPUAllocator> m_cpuOutputAllocator;
         std::shared_ptr<onnxruntime::KernelRegistry> m_kernelRegistry;
-        std::shared_ptr<const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfoMap> m_internalRegInfoMap;
+        std::shared_ptr<const Windows::AI::MachineLearning::Adapter::InternalRegistrationInfoMap> m_internalRegInfoMap;
         mutable uint64_t m_partitionKernelPrefixVal = 0;
 
         bool m_closed = false;
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h
index 9d58622657904..02319daab0ab9 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphDescBuilder.h
@@ -9,7 +9,7 @@ namespace Dml
 {
     struct GraphNodeProperties
     {
-        std::shared_ptr<const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfo> 
+        std::shared_ptr<const Windows::AI::MachineLearning::Adapter::InternalRegistrationInfo> 
             internalRegInfo;
 
         // These are currently passed from the partitioning step since the only DML operators current 
diff --git a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h
index 2227a2e67bf0a..cba44ccb6f96e 100644
--- a/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h
+++ b/onnxruntime/core/providers/dml/DmlExecutionProvider/src/GraphPartitioner.h
@@ -43,7 +43,7 @@ namespace Dml
     std::vector<std::unique_ptr<GraphPartition>>
     BuildPartitions(
         const onnxruntime::GraphViewer& graph,
-        const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfoMap& internalRegInfoMap,
+        const Windows::AI::MachineLearning::Adapter::InternalRegistrationInfoMap& internalRegInfoMap,
         const std::vector<const onnxruntime::KernelRegistry*>& registries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         std::unordered_map<const onnxruntime::Node*, GraphNodeProperties>& graphNodePropertyMap,
@@ -53,7 +53,7 @@ namespace Dml
     std::vector<std::unique_ptr<onnxruntime::ComputeCapability>>
     PartitionGraph(
         const onnxruntime::GraphViewer& graph,
-        const winrt::Windows::AI::MachineLearning::implementation::InternalRegistrationInfoMap& internalRegInfoMap,
+        const Windows::AI::MachineLearning::Adapter::InternalRegistrationInfoMap& internalRegInfoMap,
         const std::vector<const onnxruntime::KernelRegistry*>& registries,
         uint32_t supportedDeviceDataTypeMask, // Each bit corresponds to each DML_TENSOR_DATA_TYPE.
         onnxruntime::KernelRegistry* registryForPartitionKernels,
diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/SchemaInferenceOverrider.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/SchemaInferenceOverrider.h
index 19fc6280246c6..62e8c30cd19ef 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/SchemaInferenceOverrider.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/SchemaInferenceOverrider.h
@@ -80,7 +80,7 @@ OverrideSchemaInferenceFunction<OperatorHelper::ShapeInferenceHelper_##shapeInfe
         OVERRIDE_SCHEMA(    7,  true,  Crop);
         OVERRIDE_SCHEMA(    7,  false, Upsample);
         OVERRIDE_SCHEMA_EX( 9,  true,  Upsample, Resize, 1); // Upsample v9 uses Resize's shape inference function.
-        OVERRIDE_SCHEMA(    7,  true,  Slice);
+        OVERRIDE_SCHEMA_EX( 7,  true,  Slice, Slice7);
         OVERRIDE_SCHEMA(    7,  true,  Split);
         OVERRIDE_SCHEMA_EX( 7,  true,  Tile, Tile, 1);
         OVERRIDE_SCHEMA_EX( 8,  true,  Expand, Expand, 1);
diff --git a/winml/adapter/AbiCustomRegistryImpl.cpp b/winml/adapter/AbiCustomRegistryImpl.cpp
index d81512a6823f0..7242ca121ffe5 100644
--- a/winml/adapter/AbiCustomRegistryImpl.cpp
+++ b/winml/adapter/AbiCustomRegistryImpl.cpp
@@ -43,13 +43,14 @@ HRESULT STDMETHODCALLTYPE AbiCustomRegistryImpl::RegisterOperatorKernel(
     const MLOperatorKernelDescription* opKernel,
     IMLOperatorKernelFactory* operatorKernelFactory,
     _In_opt_ IMLOperatorShapeInferrer* shapeInferrer) const noexcept {
-  return RegisterOperatorKernel(opKernel, operatorKernelFactory, shapeInferrer, false, false, false);
+  return RegisterOperatorKernel(opKernel, operatorKernelFactory, shapeInferrer, nullptr, false, false, false);
 }
 
 HRESULT STDMETHODCALLTYPE AbiCustomRegistryImpl::RegisterOperatorKernel(
     const MLOperatorKernelDescription* opKernel,
     IMLOperatorKernelFactory* operatorKernelFactory,
     _In_opt_ IMLOperatorShapeInferrer* shapeInferrer,
+    _In_opt_ IMLOperatorSupportQueryPrivate* supportQuery,
     bool isInternalOperator,
     bool canAliasFirstInput,
     bool supportsGraph,
@@ -72,6 +73,7 @@ HRESULT STDMETHODCALLTYPE AbiCustomRegistryImpl::RegisterOperatorKernel(
       opKernel,
       operatorKernelFactory,
       shapeInferrer,
+      supportQuery,
       isInternalOperator,
       canAliasFirstInput,
       supportsGraph,
diff --git a/winml/adapter/AbiCustomRegistryImpl.h b/winml/adapter/AbiCustomRegistryImpl.h
index 747dd65d5a327..a07f51cacd067 100644
--- a/winml/adapter/AbiCustomRegistryImpl.h
+++ b/winml/adapter/AbiCustomRegistryImpl.h
@@ -23,6 +23,7 @@ class AbiCustomRegistryImpl : public AbiCustomRegistry {
       const MLOperatorKernelDescription* operator_kernel,
       IMLOperatorKernelFactory* operator_kernel_factory,
       _In_opt_ IMLOperatorShapeInferrer* shape_inferrer,
+      _In_opt_ IMLOperatorSupportQueryPrivate* supportQuery,
       bool is_internal_operator,
       bool can_alias_first_input,
       bool supports_graph,

From 120433c29d9ec9de7f3603e9257b15cd5c7fa08a Mon Sep 17 00:00:00 2001
From: Dmitri Smirnov <yuslepukhin@users.noreply.github.com>
Date: Mon, 13 Jan 2020 17:58:33 -0800
Subject: [PATCH 53/75] Add OneHotEncoder and HashOneHotEncoder kernels.
 (#2830)

 Add defs and imlementation for OneHotEncoders, adjuist date_time_transformer kernel and test.
  Add OneHotEncoder kernel test.
  Add HashOneHotVectorizerTransformer unit test.
  This does not link due to multiple definitions of functions
  that are included into header from a CPP file.
---
 cgmanifest.json                               |   2 +-
 cmake/external/featurizers.cmake              |   2 +-
 .../graph/featurizers_ops/featurizers_defs.cc | 367 ++++++++++++++----
 .../cpu/cat_imputer_transformer.cc            |   2 +-
 .../cpu/date_time_transformer.cc              |   7 +-
 .../hash_one_hot_vectorizer_transformer.cc    |  90 +++++
 .../cpu/imputation_marker_transformer.cc      |   4 +-
 .../cpu/label_encoder_transformer.cc          |   2 +-
 .../cpu/max_abs_scalar_transformer.cc         |   3 +-
 .../cpu/missing_dummies_transformer.cc        |   3 +-
 .../cpu/one_hot_encoder_transformer.cc        |  90 +++++
 .../cpu_featurizers_kernels.cc                |   4 +
 .../datetimetransformer_test.cc               |  37 +-
 .../hash_one_hot_encoder_transformer_test.cc  |  89 +++++
 .../featurizers_ops/one_hot_encoder_test.cc   | 119 ++++++
 15 files changed, 728 insertions(+), 93 deletions(-)
 create mode 100644 onnxruntime/featurizers_ops/cpu/hash_one_hot_vectorizer_transformer.cc
 create mode 100644 onnxruntime/featurizers_ops/cpu/one_hot_encoder_transformer.cc
 create mode 100644 onnxruntime/test/featurizers_ops/hash_one_hot_encoder_transformer_test.cc
 create mode 100644 onnxruntime/test/featurizers_ops/one_hot_encoder_test.cc

diff --git a/cgmanifest.json b/cgmanifest.json
index 331efe0d32d80..25fbd376028c7 100644
--- a/cgmanifest.json
+++ b/cgmanifest.json
@@ -450,7 +450,7 @@
       {
          "component": {
             "git": {
-               "commitHash": "573070aeeb77e267da2579ac1d75d92c688bbe97",
+               "commitHash": "ebec32ef06859b6399bf8854f18b91158c87760b",
                "repositoryUrl": "https://github.com/microsoft/FeaturizersLibrary.git"
             },
             "type": "git"
diff --git a/cmake/external/featurizers.cmake b/cmake/external/featurizers.cmake
index 4700e85f032db..5f1c49b73b2f7 100644
--- a/cmake/external/featurizers.cmake
+++ b/cmake/external/featurizers.cmake
@@ -3,7 +3,7 @@
 # This source code should not depend on the onnxruntime and may be built independently
 
 set(featurizers_URL "https://github.com/microsoft/FeaturizersLibrary.git")
-set(featurizers_TAG "573070aeeb77e267da2579ac1d75d92c688bbe97")
+set(featurizers_TAG "ebec32ef06859b6399bf8854f18b91158c87760b")
 
 set(featurizers_pref FeaturizersLibrary)
 set(featurizers_ROOT ${PROJECT_SOURCE_DIR}/external/${featurizers_pref})
diff --git a/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc b/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
index c46822774918d..6497e6cd882bd 100644
--- a/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
+++ b/onnxruntime/core/graph/featurizers_ops/featurizers_defs.cc
@@ -34,11 +34,13 @@ using ONNX_NAMESPACE::OPTIONAL;
 // Forward declarations
 static void RegisterCatImputerFeaturizerVer1();
 static void RegisterDateTimeFeaturizerVer1();
+static void RegisterHashOneHotVectorizerFeaturizerVer1();
 static void RegisterImputationMarkerFeaturizerVer1();
 static void RegisterLabelEncoderFeaturizerVer1();
 static void RegisterMaxAbsScalarFeaturizerVer1();
 static void RegisterMinMaxScalarFeaturizerVer1();
 static void RegisterMissingDummiesFeaturizerVer1();
+static void RegisterOneHotEncoderFeaturizerVer1();
 static void RegisterRobustScalarFeaturizerVer1();
 static void RegisterStringFeaturizerVer1();
 static void RegisterTimeSeriesImputerFeaturizerVer1();
@@ -49,11 +51,13 @@ static void RegisterTimeSeriesImputerFeaturizerVer1();
 void RegisterMSFeaturizersSchemas() {
   RegisterCatImputerFeaturizerVer1();
   RegisterDateTimeFeaturizerVer1();
+  RegisterHashOneHotVectorizerFeaturizerVer1();
   RegisterImputationMarkerFeaturizerVer1();
   RegisterLabelEncoderFeaturizerVer1();
   RegisterMaxAbsScalarFeaturizerVer1();
   RegisterMinMaxScalarFeaturizerVer1();
   RegisterMissingDummiesFeaturizerVer1();
+  RegisterOneHotEncoderFeaturizerVer1();
   RegisterRobustScalarFeaturizerVer1();
   RegisterStringFeaturizerVer1();
   RegisterTimeSeriesImputerFeaturizerVer1();
@@ -70,15 +74,15 @@ void RegisterCatImputerFeaturizerVer1() {
         within the host frameworks and programming languages.
 
         C++-style pseudo signature:
-          std::float_t execute(std::float_t const &value);
-          std::double_t execute(std::double_t const &value);
+          float execute(float const &value);
+          double execute(double const &value);
           template <typename T> T execute(std::optional<T> const &value);
 
         Examples (where 55.5 is the mode value):
           execute(1.0) -> 1.0
           execute(NaN) -> 55.5
           execute(2.0) -> 2.0
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(CatImputerTransformer)
       .SinceVersion(1)
@@ -121,7 +125,7 @@ void RegisterDateTimeFeaturizerVer1() {
         Extracts various datetime-related values from a UTC time_point.
 
         C++-style pseudo signature:
-          TimePoint execute(std::chron::system_clock::time_point const &value);
+          TimePoint execute(std::chrono::system_clock::time_point const &value);
 
         Examples:
           Given a time_point 'value' representing "November 17, 1976 12:27:04PM":
@@ -149,7 +153,7 @@ void RegisterDateTimeFeaturizerVer1() {
             "holidayName": "",
             "isPaidTimeOff": 0
           }
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(DateTimeTransformer)
       .SinceVersion(1)
@@ -209,52 +213,191 @@ void RegisterDateTimeFeaturizerVer1() {
       .TypeAndShapeInferenceFunction(
           [](ONNX_NAMESPACE::InferenceContext& ctx) {
             const bool has_shape = hasInputShape(ctx, 1);
-            for (int output = 0; output < 21; ++output) {
-              switch (output) {
-                case 0:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, output);
-                  break;
-                case 1:  // fall through
-                case 2:
-                case 3:
-                case 4:
-                case 5:
-                case 6:
-                case 7:
-                case 8:
-                case 9:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
-                  break;
-                case 10:  // fall through
-                case 11:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT16, output);
-                  break;
-                case 12:  // fall through
-                case 13:
-                case 14:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
-                  break;
-                case 15:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, output);
-                  break;
-                case 16:
-                case 17:
-                case 18:
-                case 19:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, output);
-                  break;
-                case 20:
-                  propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, output);
-                  break;
-                default:
-                  assert(false);
-                  break;
-              }
-              if (has_shape) {
-                propagateShapeFromInputToOutput(ctx, 1, output);
-              }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, 0);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
             }
-          });
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 1);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 1);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 2);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 2);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 3);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 3);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 4);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 4);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 5);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 5);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 6);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 6);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 7);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 7);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 8);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 8);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 9);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 9);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT16, 10);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 10);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT16, 11);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 11);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 12);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 12);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 13);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 13);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 14);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 14);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_INT32, 15);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 15);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, 16);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 16);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, 17);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 17);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, 18);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 18);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_STRING, 19);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 19);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 20);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 20);
+            }
+
+          }
+      );
+}
+
+void RegisterHashOneHotVectorizerFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        Hashes the input to a categorical value, then produces a one hot encoded vector
+        based on that value.
+
+        C++-style pseudo signature:
+            template <typename T> HashOneHotVectorizerStruct execute(T const &value);
+
+        Examples:
+          Assuming the hashing algorithm...
+            "A" -> 1
+            "B" -> 2
+            "C" -> 5
+
+          and 'numCols' set to 8:
+
+            execute("A") -> [1, 0, 0, 0, 0, 0, 0, 0]
+            execute("B") -> [0, 1, 0, 0, 0, 0, 0, 0]
+            execute("C") -> [0, 0, 0, 0, 1, 0, 0, 0]
+  )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(HashOneHotVectorizerTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(0, "NumElements", "No information available", "OutputT0")
+      .Output(1, "Value", "No information available", "OutputT1")
+      .Output(2, "Index", "No information available", "OutputT0")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(int8)", "tensor(int16)", "tensor(int32)", "tensor(int64)", "tensor(uint8)", "tensor(uint16)", "tensor(uint32)", "tensor(uint64)", "tensor(float)", "tensor(double)", "tensor(bool)", "tensor(string)"},
+          "No information is available")
+      .TypeConstraint(
+          "OutputT0",
+          {"tensor(uint64)"},
+          "No information is available")
+      .TypeConstraint(
+          "OutputT1",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            const bool has_shape = hasInputShape(ctx, 1);
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT64, 0);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 1);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 1);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT64, 2);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 2);
+            }
+
+          }
+      );
 }
 
 void RegisterImputationMarkerFeaturizerVer1() {
@@ -262,17 +405,17 @@ void RegisterImputationMarkerFeaturizerVer1() {
         Returns true if the input is null, false if it is not.
 
         C++-style pseudo signature:
-          bool execute(std::float_t const &value);
-          bool execute(std::double_t const &value);
+          bool execute(float const &value);
+          bool execute(double const &value);
           template <typename T> bool execute(std::optional<T> const &value);
 
         Examples:
           3.0 -> false
           NaN -> true
           "foo" -> false
-          std::optional<std::string>() -> true
-          std::optional<std::string>("bar") -> false
-    )DOC";
+          std::optional<string>() -> true
+          std::optional<string>("bar") -> false
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(ImputationMarkerTransformer)
       .SinceVersion(1)
@@ -315,7 +458,7 @@ void RegisterLabelEncoderFeaturizerVer1() {
         Returns a unique id for the input based on all values encountered during training.
 
         C++-style pseudo signature:
-          template <typename T> std::uint32_t execute(T const &value);
+          template <typename T> uint32 execute(T const &value);
 
         Examples:
           Assuming the training data of ["A", "B", "C"]...
@@ -324,7 +467,7 @@ void RegisterLabelEncoderFeaturizerVer1() {
           execute("B") -> 2
           execute("C") -> 3
           execute("This value was not seen during training") -> 0
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(LabelEncoderTransformer)
       .SinceVersion(1)
@@ -367,8 +510,8 @@ void RegisterMaxAbsScalarFeaturizerVer1() {
         Scales input based on the maximum absolute value of all data encountered during training.
 
         C++-style pseudo signature:
-          std::float_t execute(std::uint16_t value);
-          std::double_t execute(std::uint32_t value);
+          float execute(uint16 value);
+          double execute(uint32 value);
 
         Examples:
           Given a training set of [1.0, -2.0, 3.0, -4.0], where 4.0 is the absolute value of the
@@ -377,7 +520,7 @@ void RegisterMaxAbsScalarFeaturizerVer1() {
           execute(1.0) -> 1.0 / 4.0
           execute(-4.0) -> -4.0 / 4.0
           execute(100.0) -> 100 / 4.0
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(MaxAbsScalarTransformer)
       .SinceVersion(1)
@@ -426,11 +569,13 @@ void RegisterMaxAbsScalarFeaturizerVer1() {
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_DOUBLE) {
               propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_DOUBLE, 0);
             } else {
-              fail_type_inference("input 1 is expected to have a accepted type");
+              fail_type_inference("input 1 is expected to have an accepted type");
             }
+
             if (hasInputShape(ctx, 1)) {
               propagateShapeFromInputToOutput(ctx, 1, 0);
             }
+
           });
 }
 
@@ -440,7 +585,7 @@ void RegisterMinMaxScalarFeaturizerVer1() {
         during training.
 
         C++-style pseudo signature:
-            template <typeanem T> std::double_t(T const &value);
+            template <typeanem T> double(T const &value);
 
         Examples:
           Given the training data [1, 2, 3, 4, 5];
@@ -450,7 +595,7 @@ void RegisterMinMaxScalarFeaturizerVer1() {
 
           execute(2) = 2 / 4
           execute(20) = 20 / 4
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(MinMaxScalarTransformer)
       .SinceVersion(1)
@@ -493,17 +638,17 @@ void RegisterMissingDummiesFeaturizerVer1() {
         Returns 1 if the input is null, 0 if it is not.
 
         C++-style pseudo signature:
-            std::int8_t execute(std::float_t const &value);
-            std::int8_t execute(std::double_t const &value);
-            template <typename T> std::int8_t execute(T const &value);
+            int8 execute(float const &value);
+            int8 execute(double const &value);
+            template <typename T> int8 execute(T const &value);
 
         Examples:
           1.0 -> 0
           NaN -> 1
           "foo" -> 0
-          std::optional<std::string>() -> 1
-          std::optional<std::string>("bar") -> 0
-    )DOC";
+          std::optional<string>() -> 1
+          std::optional<string>("bar") -> 0
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(MissingDummiesTransformer)
       .SinceVersion(1)
@@ -541,6 +686,80 @@ void RegisterMissingDummiesFeaturizerVer1() {
           });
 }
 
+void RegisterOneHotEncoderFeaturizerVer1() {
+  static const char* doc = R"DOC(
+        Produces a one hot vector based on categories calculated during training.
+
+        C++-style pseudo signature:
+          template <typename T> OneHotVector execute(T const &value);
+
+        Examples:
+          Assuming the training data [10, 20, 30, 40]...
+
+          execute(10) -> [0, 1, 0, 0, 0]
+          execute(20) -> [0, 0, 1, 0, 0]
+          execute(30) -> [0, 0, 0, 1, 0]
+          execute(40) -> [0, 0, 0, 0, 1]
+          execute(200) -> [1, 0, 0, 0, 0]
+          execute(-1) -> [1, 0, 0, 0, 0]
+  )DOC";
+
+  MS_FEATURIZERS_OPERATOR_SCHEMA(OneHotEncoderTransformer)
+      .SinceVersion(1)
+      .SetDomain(kMSFeaturizersDomain)
+      .SetDoc(doc)
+      .Input(
+          0,
+          "State",
+          "State generated during training that is used for prediction",
+          "T0")
+      .Input(
+          1,
+          "Input",
+          "No information is available",
+          "InputT")
+      .Output(0, "NumElements", "No information available", "OutputT0")
+      .Output(1, "Value", "No information available", "OutputT1")
+      .Output(2, "Index", "No information available", "OutputT0")
+      .TypeConstraint(
+          "T0",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeConstraint(
+          "InputT",
+          {"tensor(int8)", "tensor(int16)", "tensor(int32)", "tensor(int64)", "tensor(uint8)", "tensor(uint16)", "tensor(uint32)", "tensor(uint64)", "tensor(float)", "tensor(double)", "tensor(bool)", "tensor(string)"},
+          "No information is available")
+      .TypeConstraint(
+          "OutputT0",
+          {"tensor(uint64)"},
+          "No information is available")
+      .TypeConstraint(
+          "OutputT1",
+          {"tensor(uint8)"},
+          "No information is available")
+      .TypeAndShapeInferenceFunction(
+          [](ONNX_NAMESPACE::InferenceContext& ctx) {
+            const bool has_shape = hasInputShape(ctx, 1);
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT64, 0);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 0);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT8, 1);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 1);
+            }
+
+            propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_UINT64, 2);
+            if(has_shape) {
+              propagateShapeFromInputToOutput(ctx, 1, 2);
+            }
+
+          }
+      );
+}
+
 void RegisterRobustScalarFeaturizerVer1() {
   static const char* doc = R"DOC(
         MinMaxScalarEstimator + centering?
@@ -550,7 +769,7 @@ void RegisterRobustScalarFeaturizerVer1() {
 
         Examples:
           TODO
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(RobustScalarTransformer)
       .SinceVersion(1)
@@ -599,11 +818,13 @@ void RegisterRobustScalarFeaturizerVer1() {
                        input_elem_type == ONNX_NAMESPACE::TensorProto_DataType_DOUBLE) {
               propagateElemTypeFromDtypeToOutput(ctx, ONNX_NAMESPACE::TensorProto_DataType_DOUBLE, 0);
             } else {
-              fail_type_inference("input 1 is expected to have a accepted type");
+              fail_type_inference("input 1 is expected to have an accepted type");
             }
+
             if (hasInputShape(ctx, 1)) {
               propagateShapeFromInputToOutput(ctx, 1, 0);
             }
+
           });
 }
 
@@ -612,12 +833,12 @@ void RegisterStringFeaturizerVer1() {
         Converts the input into a string representation based on the input's type.
 
         C++-style pseudo signature:
-          template <typename T> std::string execute(T const &value);
+          template <typename T> string execute(T const &value);
 
         Examples:
           execute(1) -> "1"
           execute(3.14) -> "3.14"
-    )DOC";
+  )DOC";
 
   MS_FEATURIZERS_OPERATOR_SCHEMA(StringTransformer)
       .SinceVersion(1)
diff --git a/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc b/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc
index b2adb9108f8b9..7a5ee782e3d2f 100644
--- a/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/cat_imputer_transformer.cc
@@ -27,7 +27,7 @@ inline nonstd::optional<std::string> PreprocessOptional(std::string value) {
   return value.empty() ? nonstd::optional<std::string>() : nonstd::optional<std::string>(std::move(value));
 }
 
-template<typename T>
+template <typename T>
 struct CatImputerTransformerImpl {
   void operator()(OpKernelContext* ctx) const {
     // Create the transformer
diff --git a/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc b/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc
index 0e25e58fac68e..13f780414d988 100644
--- a/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/date_time_transformer.cc
@@ -3,6 +3,7 @@
 
 #include "core/common/common.h"
 #include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
 #include "core/framework/op_kernel.h"
 
 #include "Featurizers/DateTimeFeaturizer.h"
@@ -29,7 +30,7 @@ class DateTimeTransformer final : public OpKernel {
 
     // Get the input
     const auto* input_tensor(ctx->Input<Tensor>(1));
-    const std::int64_t* input_data(input_tensor->Data<std::int64_t>());
+    const int64_t* input_data(input_tensor->Data<int64_t>());
 
     // Prepare the output
     Tensor* year_tensor(ctx->Output(0, input_tensor->Shape()));
@@ -80,7 +81,7 @@ class DateTimeTransformer final : public OpKernel {
     const int64_t length(input_tensor->Shape().Size());
 
     for (int64_t i = 0; i < length; ++i) {
-      auto result(transformer.execute(input_data[i]));
+      auto result(transformer.execute(std::chrono::system_clock::from_time_t(input_data[i])));
 
       year_data[i] = std::move(result.year);
       month_data[i] = std::move(result.month);
@@ -116,7 +117,7 @@ ONNX_OPERATOR_KERNEL_EX(
     kCpuExecutionProvider,
     KernelDefBuilder()
         .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
-        .TypeConstraint("T1", DataTypeImpl::GetTensorType<uint64_t>()),
+        .TypeConstraint("T1", DataTypeImpl::GetTensorType<int64_t>()),
     DateTimeTransformer);
 
 }  // namespace featurizers
diff --git a/onnxruntime/featurizers_ops/cpu/hash_one_hot_vectorizer_transformer.cc b/onnxruntime/featurizers_ops/cpu/hash_one_hot_vectorizer_transformer.cc
new file mode 100644
index 0000000000000..747dcf885cb23
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/hash_one_hot_vectorizer_transformer.cc
@@ -0,0 +1,90 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/HashOneHotVectorizerFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+template <typename InputT>
+struct HashOneHotVectorizerTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::HashOneHotVectorizerTransformer<InputT> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::HashOneHotVectorizerTransformer<InputT>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* NumElements_tensor(ctx->Output(0, input_tensor->Shape()));
+    Tensor* Value_tensor(ctx->Output(1, input_tensor->Shape()));
+    Tensor* Index_tensor(ctx->Output(2, input_tensor->Shape()));
+
+    uint64_t* NumElements_data(NumElements_tensor->MutableData<uint64_t>());
+    uint8_t* Value_data(Value_tensor->MutableData<uint8_t>());
+    uint64_t* Index_data(Index_tensor->MutableData<uint64_t>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      auto result(transformer.execute(input_data[i]));
+
+      NumElements_data[i] = std::move(result.NumElements);
+      Value_data[i] = std::move(result.Value);
+      Index_data[i] = std::move(result.Index);
+    }
+  }
+};
+
+class HashOneHotVectorizerTransformer final : public OpKernel {
+ public:
+  explicit HashOneHotVectorizerTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<HashOneHotVectorizerTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t,
+                                uint32_t, int64_t, uint64_t, float, double, bool, std::string>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    HashOneHotVectorizerTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>(),
+                                   DataTypeImpl::GetTensorType<bool>(),
+                                   DataTypeImpl::GetTensorType<std::string>()}),
+    HashOneHotVectorizerTransformer);
+
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc b/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc
index a0f36a943ff01..2612f448e7335 100644
--- a/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/imputation_marker_transformer.cc
@@ -54,8 +54,7 @@ class ImputationMarkerTransformer final : public OpKernel {
   }
 
   Status Compute(OpKernelContext* ctx) const override {
-    utils::MLTypeCallDispatcher<ImputationMarkerTransformerImpl, float, double, std::string>
-        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    utils::MLTypeCallDispatcher<ImputationMarkerTransformerImpl, float, double, std::string> t_disp(ctx->Input<Tensor>(1)->GetElementType());
     t_disp.Invoke(ctx);
     return Status::OK();
   }
@@ -72,5 +71,6 @@ ONNX_OPERATOR_KERNEL_EX(
                                    DataTypeImpl::GetTensorType<double>(),
                                    DataTypeImpl::GetTensorType<std::string>()}),
     ImputationMarkerTransformer);
+
 }  // namespace featurizers
 }  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc b/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc
index 96ec25285133c..6350da43fc906 100644
--- a/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/label_encoder_transformer.cc
@@ -31,7 +31,7 @@ struct LabelEncoderTransformerImpl {
 
     // Prepare the output
     Tensor* output_tensor(ctx->Output(0, input_tensor->Shape()));
-    std::uint32_t* output_data(output_tensor->MutableData<std::uint32_t>());
+    uint32_t* output_data(output_tensor->MutableData<uint32_t>());
 
     // Execute
     const int64_t length(input_tensor->Shape().Size());
diff --git a/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc b/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc
index 8fea4f9fb806e..0208db93af0bf 100644
--- a/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/max_abs_scalar_transformer.cc
@@ -72,7 +72,8 @@ class MaxAbsScalarTransformer final : public OpKernel {
 
   Status Compute(OpKernelContext* ctx) const override {
     utils::MLTypeCallDispatcher<MaxAbsScalarTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
-      int64_t, uint64_t, float, double> t_disp(ctx->Input<Tensor>(1)->GetElementType());
+                                int64_t, uint64_t, float, double>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
     t_disp.Invoke(ctx);
     return Status::OK();
   }
diff --git a/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc b/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc
index 255bfd78d3cc3..5d40abcf761ad 100644
--- a/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc
+++ b/onnxruntime/featurizers_ops/cpu/missing_dummies_transformer.cc
@@ -54,8 +54,7 @@ class MissingDummiesTransformer final : public OpKernel {
   }
 
   Status Compute(OpKernelContext* ctx) const override {
-    utils::MLTypeCallDispatcher<MissingDummiesTransformerImpl, float, double, std::string>
-        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    utils::MLTypeCallDispatcher<MissingDummiesTransformerImpl, float, double, std::string> t_disp(ctx->Input<Tensor>(1)->GetElementType());
     t_disp.Invoke(ctx);
     return Status::OK();
   }
diff --git a/onnxruntime/featurizers_ops/cpu/one_hot_encoder_transformer.cc b/onnxruntime/featurizers_ops/cpu/one_hot_encoder_transformer.cc
new file mode 100644
index 0000000000000..1a299456ecbed
--- /dev/null
+++ b/onnxruntime/featurizers_ops/cpu/one_hot_encoder_transformer.cc
@@ -0,0 +1,90 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "core/common/common.h"
+#include "core/framework/data_types.h"
+#include "core/framework/data_types_internal.h"
+#include "core/framework/op_kernel.h"
+
+#include "Featurizers/OneHotEncoderFeaturizer.h"
+#include "Archive.h"
+
+namespace onnxruntime {
+namespace featurizers {
+
+template <typename InputT>
+struct OneHotEncoderTransformerImpl {
+  void operator()(OpKernelContext* ctx) const {
+    // Create the transformer
+    Microsoft::Featurizer::Featurizers::OneHotEncoderTransformer<InputT> transformer(
+        [ctx](void) {
+          const auto* state_tensor(ctx->Input<Tensor>(0));
+          const uint8_t* const state_data(state_tensor->Data<uint8_t>());
+
+          Microsoft::Featurizer::Archive archive(state_data, state_tensor->Shape().GetDims()[0]);
+          return Microsoft::Featurizer::Featurizers::OneHotEncoderTransformer<InputT>(archive);
+        }());
+
+    // Get the input
+    const auto* input_tensor(ctx->Input<Tensor>(1));
+    const InputT* input_data(input_tensor->Data<InputT>());
+
+    // Prepare the output
+    Tensor* NumElements_tensor(ctx->Output(0, input_tensor->Shape()));
+    Tensor* Value_tensor(ctx->Output(1, input_tensor->Shape()));
+    Tensor* Index_tensor(ctx->Output(2, input_tensor->Shape()));
+
+    uint64_t* NumElements_data(NumElements_tensor->MutableData<uint64_t>());
+    uint8_t* Value_data(Value_tensor->MutableData<uint8_t>());
+    uint64_t* Index_data(Index_tensor->MutableData<uint64_t>());
+
+    // Execute
+    const int64_t length(input_tensor->Shape().Size());
+
+    for (int64_t i = 0; i < length; ++i) {
+      auto result(transformer.execute(input_data[i]));
+
+      NumElements_data[i] = std::move(result.NumElements);
+      Value_data[i] = std::move(result.Value);
+      Index_data[i] = std::move(result.Index);
+    }
+  }
+};
+
+class OneHotEncoderTransformer final : public OpKernel {
+ public:
+  explicit OneHotEncoderTransformer(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* ctx) const override {
+    utils::MLTypeCallDispatcher<OneHotEncoderTransformerImpl, int8_t, uint8_t, int16_t, uint16_t, int32_t, uint32_t,
+                                int64_t, uint64_t, float, double, bool, std::string>
+        t_disp(ctx->Input<Tensor>(1)->GetElementType());
+    t_disp.Invoke(ctx);
+    return Status::OK();
+  }
+};
+
+ONNX_OPERATOR_KERNEL_EX(
+    OneHotEncoderTransformer,
+    kMSFeaturizersDomain,
+    1,
+    kCpuExecutionProvider,
+    KernelDefBuilder()
+        .TypeConstraint("T0", DataTypeImpl::GetTensorType<uint8_t>())
+        .TypeConstraint("InputT", {DataTypeImpl::GetTensorType<int8_t>(),
+                                   DataTypeImpl::GetTensorType<uint8_t>(),
+                                   DataTypeImpl::GetTensorType<int16_t>(),
+                                   DataTypeImpl::GetTensorType<uint16_t>(),
+                                   DataTypeImpl::GetTensorType<int32_t>(),
+                                   DataTypeImpl::GetTensorType<uint32_t>(),
+                                   DataTypeImpl::GetTensorType<int64_t>(),
+                                   DataTypeImpl::GetTensorType<uint64_t>(),
+                                   DataTypeImpl::GetTensorType<float>(),
+                                   DataTypeImpl::GetTensorType<double>(),
+                                   DataTypeImpl::GetTensorType<bool>(),
+                                   DataTypeImpl::GetTensorType<std::string>()}),
+    OneHotEncoderTransformer);
+
+}  // namespace featurizers
+}  // namespace onnxruntime
diff --git a/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc b/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
index 2fa7168880743..96326b10de936 100644
--- a/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
+++ b/onnxruntime/featurizers_ops/cpu_featurizers_kernels.cc
@@ -12,11 +12,13 @@ namespace featurizers {
 // Forward declarations
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, CatImputerTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, DateTimeTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, HashOneHotVectorizerTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, ImputationMarkerTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, LabelEncoderTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MaxAbsScalarTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MinMaxScalarTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MissingDummiesTransformer);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, OneHotEncoderTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, RobustScalarTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, StringTransformer);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, TimeSeriesImputerTransformer);
@@ -25,11 +27,13 @@ Status RegisterCpuMSFeaturizersKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, CatImputerTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, DateTimeTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, HashOneHotVectorizerTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, ImputationMarkerTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, LabelEncoderTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MaxAbsScalarTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MinMaxScalarTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, MissingDummiesTransformer)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, OneHotEncoderTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, RobustScalarTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, StringTransformer)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSFeaturizersDomain, 1, TimeSeriesImputerTransformer)>,
diff --git a/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc b/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc
index 1044667fd54be..354f201661dee 100644
--- a/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc
+++ b/onnxruntime/test/featurizers_ops/datetimetransformer_test.cc
@@ -4,6 +4,7 @@
 #include "gtest/gtest.h"
 #include "test/providers/provider_test_utils.h"
 
+#include "Archive.h"
 #include "Featurizers/DateTimeFeaturizer.h"
 
 namespace dft = Microsoft::Featurizer::Featurizers;
@@ -13,12 +14,21 @@ using SysClock = std::chrono::system_clock;
 namespace onnxruntime {
 namespace test {
 
+std::vector<uint8_t> GetStream () {
+  dft::DateTimeTransformer dt("", "");
+  Microsoft::Featurizer::Archive ar;
+  dt.save(ar);
+  return ar.commit();
+}
+
 TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17_12_27_04) {
   const time_t date = 217081624;
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // Add state input
-  test.AddInput<uint8_t>("State", {8}, {1, 0, 0, 0, 0, 0, 0, 0});
+  auto stream = GetStream();
+  auto dim = static_cast<int64_t>(stream.size());
+  test.AddInput<uint8_t>("State", {dim}, stream);
 
   // We are adding a scalar Tensor in this instance
   test.AddInput<int64_t>("Date", {1}, {date});
@@ -76,17 +86,20 @@ TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17_12_27_04) {
 
 TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17_12_27_05) {
   const time_t date = 217081625;
+  const auto date_tp = SysClock::from_time_t(date);
 
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // Add state input
-  test.AddInput<uint8_t>("State", {8}, {1, 0, 0, 0, 0, 0, 0, 0});
+  auto stream = GetStream();
+  auto dim = static_cast<int64_t>(stream.size());
+  test.AddInput<uint8_t>("State", {dim}, stream);
 
   // We are adding a scalar Tensor in this instance
   test.AddInput<int64_t>("Date", {1}, {date});
 
   dft::DateTimeTransformer dt("", "");
-  dft::TimePoint tp(dt.execute(date));
+  dft::TimePoint tp(dt.execute(date_tp));
   ASSERT_EQ(tp.year, 1976);
   ASSERT_EQ(tp.month, dft::TimePoint::NOVEMBER);
   ASSERT_EQ(tp.day, 17);
@@ -137,19 +150,24 @@ TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17_12_27_05) {
 
 TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17__12_27_05_and_past_1976_nov_17_12_27_04) {
   const time_t date1 = 217081625;
+  const auto date1_tp = SysClock::from_time_t(date1);
   const time_t date2 = 217081624;
+  const auto date2_tp = SysClock::from_time_t(date2);
 
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // Add state input
-  test.AddInput<uint8_t>("State", {8}, {1, 0, 0, 0, 0, 0, 0, 0});
+  auto stream = GetStream();
+  auto dim = static_cast<int64_t>(stream.size());
+  test.AddInput<uint8_t>("State", {dim}, stream);
+
 
   // We are adding a scalar Tensor in this instance
   test.AddInput<int64_t>("Date", {2}, {date1, date2});
 
   dft::DateTimeTransformer dt("", "");
-  dft::TimePoint tp1(dt.execute(date1));
-  dft::TimePoint tp2(dt.execute(date2));
+  dft::TimePoint tp1(dt.execute(date1_tp));
+  dft::TimePoint tp2(dt.execute(date2_tp));
 
   // Date1
   ASSERT_EQ(tp1.year, 1976);
@@ -225,17 +243,20 @@ TEST(FeaturizersTests, DateTimeTransformer_past_1976_nov_17__12_27_05_and_past_1
 
 TEST(FeaturizersTests, DateTimeTransformer_future_2025_june_30) {
   const time_t date = 1751241600;
+  const auto date_tp = std::chrono::system_clock::from_time_t(date);
 
   OpTester test("DateTimeTransformer", 1, onnxruntime::kMSFeaturizersDomain);
 
   // Add state input
-  test.AddInput<uint8_t>("State", {8}, {1, 0, 0, 0, 0, 0, 0, 0});
+  auto stream = GetStream();
+  auto dim = static_cast<int64_t>(stream.size());
+  test.AddInput<uint8_t>("State", {dim}, stream);
 
   // We are adding a scalar Tensor in this instance
   test.AddInput<int64_t>("Date", {1}, {date});
 
   dft::DateTimeTransformer dt("", "");
-  dft::TimePoint tp = dt.execute(date);
+  dft::TimePoint tp = dt.execute(date_tp);
   ASSERT_EQ(tp.year, 2025);
   ASSERT_EQ(tp.month, dft::TimePoint::JUNE);
   ASSERT_EQ(tp.day, 30);
diff --git a/onnxruntime/test/featurizers_ops/hash_one_hot_encoder_transformer_test.cc b/onnxruntime/test/featurizers_ops/hash_one_hot_encoder_transformer_test.cc
new file mode 100644
index 0000000000000..915173110b69c
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/hash_one_hot_encoder_transformer_test.cc
@@ -0,0 +1,89 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Featurizers/HashOneHotVectorizerFeaturizer.h"
+#include "Featurizers/TestHelpers.h"
+#include "Archive.h"
+
+namespace NS = Microsoft::Featurizer;
+
+namespace onnxruntime {
+namespace test {
+
+template <typename T>
+std::vector<uint8_t> GetStream() {
+  NS::Featurizers::HashOneHotVectorizerTransformer<T> hvtransformer(2, 100);
+  NS::Archive ar;
+  hvtransformer.save(ar);
+  return ar.commit();
+}
+
+TEST(FeaturizersTests, HashOneHotVectorizerTransformer_int8) {
+  using Type = int8_t;
+  auto stream = GetStream<Type>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("HashOneHotVectorizerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<Type>("Input", {1}, {15});
+
+  test.AddOutput<uint64_t>("NumElements", {1}, {100u});
+  test.AddOutput<uint8_t>("Value", {1}, {1u});
+  test.AddOutput<uint64_t>("Index", {1}, {29u});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, HashOneHotVectorizerTransformer_int32) {
+  using Type = int32_t;
+  auto stream = GetStream<Type>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("HashOneHotVectorizerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<Type>("Input", {1}, {15});
+
+  test.AddOutput<uint64_t>("NumElements", {1}, {100u});
+  test.AddOutput<uint8_t>("Value", {1}, {1u});
+  test.AddOutput<uint64_t>("Index", {1}, {22u});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, HashOneHotVectorizerTransformer_double) {
+  using Type = double;
+  auto stream = GetStream<Type>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("HashOneHotVectorizerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<Type>("Input", {1}, {15.0});
+
+  test.AddOutput<uint64_t>("NumElements", {1}, {100u});
+  test.AddOutput<uint8_t>("Value", {1}, {1u});
+  test.AddOutput<uint64_t>("Index", {1}, {99u});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, HashOneHotVectorizerTransformer_string) {
+  using Type = std::string;
+  auto stream = GetStream<Type>();
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("HashOneHotVectorizerTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<Type>("Input", {1}, {"hello"});
+
+  test.AddOutput<uint64_t>("NumElements", {1}, {100u});
+  test.AddOutput<uint8_t>("Value", {1}, {1u});
+  test.AddOutput<uint64_t>("Index", {1}, {25u});
+
+  test.Run();
+}
+
+}  // namespace test
+}  // namespace onnxruntime
\ No newline at end of file
diff --git a/onnxruntime/test/featurizers_ops/one_hot_encoder_test.cc b/onnxruntime/test/featurizers_ops/one_hot_encoder_test.cc
new file mode 100644
index 0000000000000..f60d0b2d60ce2
--- /dev/null
+++ b/onnxruntime/test/featurizers_ops/one_hot_encoder_test.cc
@@ -0,0 +1,119 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#include "gtest/gtest.h"
+#include "test/providers/provider_test_utils.h"
+
+#include "Featurizers/OneHotEncoderFeaturizer.h"
+#include "Featurizers/TestHelpers.h"
+#include "Archive.h"
+
+namespace NS = Microsoft::Featurizer;
+
+namespace onnxruntime {
+namespace test {
+
+template <typename InputType>
+std::vector<uint8_t> GetStream(const std::vector<std::vector<InputType>>& trainingBatches, bool allowMissingValues) {
+  using Estimator = NS::Featurizers::OneHotEncoderEstimator<InputType>;
+
+  Estimator estimator(NS::CreateTestAnnotationMapsPtr(1), 0, allowMissingValues);
+  NS::TestHelpers::Train<Estimator, InputType>(estimator, trainingBatches);
+  typename Estimator::TransformerUniquePtr pTransformer(estimator.create_transformer());
+
+  NS::Archive ar;
+  pTransformer->save(ar);
+  return ar.commit();
+}
+
+TEST(FeaturizersTests, OneHotEncoder_uint32_t) {
+  using InputType = uint32_t;
+
+  auto trainingBatches = NS::TestHelpers::make_vector<std::vector<InputType>>(
+      NS::TestHelpers::make_vector<InputType>(10, 20, 10),
+      NS::TestHelpers::make_vector<InputType>(30),
+      NS::TestHelpers::make_vector<InputType>(10, 10, 11, 15),
+      NS::TestHelpers::make_vector<InputType>(18, 8));
+
+  auto stream = GetStream<InputType>(trainingBatches, false);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("OneHotEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {5}, {11u, 8u, 10u, 15u, 20u});
+
+  test.AddOutput<uint64_t>("NumElements", {5}, {7u, 7u, 7u, 7u, 7u});
+  test.AddOutput<uint8_t>("Value", {5}, {1u, 1u, 1u, 1u, 1u});
+  test.AddOutput<uint64_t>("Index", {5}, {2u, 0u, 1u, 3u, 5u});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, OneHotEncoder_string) {
+  using InputType = std::string;
+
+  auto trainingBatches = NS::TestHelpers::make_vector<std::vector<InputType>>(
+      NS::TestHelpers::make_vector<InputType>("orange", "apple", "orange",
+                                              "grape", "carrot", "carrot",
+                                              "peach", "banana", "orange"));
+
+  auto stream = GetStream<InputType>(trainingBatches, false);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("OneHotEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {4}, {"banana", "grape", "apple", "orange"});
+
+  test.AddOutput<uint64_t>("NumElements", {4}, {6u, 6u, 6u, 6u});
+  test.AddOutput<uint8_t>("Value", {4}, {1u, 1u, 1u, 1u});
+  test.AddOutput<uint64_t>("Index", {4}, {1u, 3u, 0u, 4u});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, OneHotEncoder_unseen_values) {
+  using InputType = std::string;
+
+  auto trainingBatches = NS::TestHelpers::make_vector<std::vector<InputType>>(
+      NS::TestHelpers::make_vector<InputType>("orange", "apple", "orange",
+                                              "grape", "carrot", "carrot",
+                                              "peach", "banana", "orange"));
+
+  auto stream = GetStream<InputType>(trainingBatches, true);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("OneHotEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {5}, {"banana", "grape", "apple", "orange", "hello"});
+
+  test.AddOutput<uint64_t>("NumElements", {5}, {7u, 7u, 7u, 7u, 7u});
+  test.AddOutput<uint8_t>("Value", {5}, {1u, 1u, 1u, 1u, 1u});
+  test.AddOutput<uint64_t>("Index", {5}, {2u, 4u, 1u, 5u, 0u});
+
+  test.Run();
+}
+
+TEST(FeaturizersTests, OneHotEncoder_unseen_values_throws) {
+  using InputType = std::string;
+
+  auto trainingBatches = NS::TestHelpers::make_vector<std::vector<InputType>>(
+      NS::TestHelpers::make_vector<InputType>("orange", "apple", "orange",
+                                              "grape", "carrot", "carrot",
+                                              "peach", "banana", "orange"));
+
+  auto stream = GetStream<InputType>(trainingBatches, false);
+  auto dim = static_cast<int64_t>(stream.size());
+
+  OpTester test("OneHotEncoderTransformer", 1, onnxruntime::kMSFeaturizersDomain);
+  test.AddInput<uint8_t>("State", {dim}, stream);
+  test.AddInput<InputType>("Input", {5}, {"banana", "grape", "apple", "orange", "hello"});
+
+  test.AddOutput<uint64_t>("NumElements", {5}, {7u, 7u, 7u, 7u, 7u});
+  test.AddOutput<uint8_t>("Value", {5}, {1u, 1u, 1u, 1u, 1u});
+  test.AddOutput<uint64_t>("Index", {5}, {2u, 4u, 1u, 5u, 0u});
+
+  test.Run(OpTester::ExpectResult::kExpectFailure);
+}
+
+}  // namespace test
+}  // namespace onnxruntime

From 5c391854f468d97ec576b6b879a77a36cd8a4bec Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Mon, 13 Jan 2020 20:16:48 -0800
Subject: [PATCH 54/75] Upgrade gtest to the latest version (#2827)

WinML would like to update the googletest submodule. They want some newer features (namely GTEST_SKIP to skip tests programmatically and be able to skip entire fixtures easily) and would need to update the submodule version.

However, because the new version of code hit a bug in gcc, even though the bug is already fixed in the latest gcc but we're using gcc 4.8.x and it won't get patched for the bug, so we have to do a compromise, change our code a little bit to make it work.

The gcc bug:  https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
---
 cgmanifest.json                                          | 2 +-
 cmake/external/googletest                                | 2 +-
 onnxruntime/test/framework/execution_frame_test.cc       | 9 ++++++---
 onnxruntime/test/framework/float_16_test.cc              | 3 ++-
 onnxruntime/test/framework/inference_session_test.cc     | 9 ++++++---
 onnxruntime/test/framework/local_kernel_registry_test.cc | 3 ++-
 onnxruntime/test/framework/tensor_test.cc                | 9 ++++++---
 onnxruntime/test/optimizer/graph_transform_test.cc       | 3 ++-
 onnxruntime/test/providers/cpu/controlflow/loop_test.cc  | 3 ++-
 9 files changed, 28 insertions(+), 15 deletions(-)

diff --git a/cgmanifest.json b/cgmanifest.json
index 25fbd376028c7..b76d46cc434c2 100644
--- a/cgmanifest.json
+++ b/cgmanifest.json
@@ -22,7 +22,7 @@
          "component": {
             "type": "git",
             "git": {
-               "commitHash": "9bda90b7e5e08c4c37a832d0cea218aed6af6470",
+               "commitHash": "703bd9caab50b139428cea1aaff9974ebee5742e",
                "repositoryUrl": "https://github.com/google/googletest.git"
             }
          }
diff --git a/cmake/external/googletest b/cmake/external/googletest
index 9bda90b7e5e08..703bd9caab50b 160000
--- a/cmake/external/googletest
+++ b/cmake/external/googletest
@@ -1 +1 @@
-Subproject commit 9bda90b7e5e08c4c37a832d0cea218aed6af6470
+Subproject commit 703bd9caab50b139428cea1aaff9974ebee5742e
diff --git a/onnxruntime/test/framework/execution_frame_test.cc b/onnxruntime/test/framework/execution_frame_test.cc
index ee67b5b473a97..36fcc9c1cbf8e 100644
--- a/onnxruntime/test/framework/execution_frame_test.cc
+++ b/onnxruntime/test/framework/execution_frame_test.cc
@@ -90,7 +90,8 @@ TEST_F(ExecutionFrameTest, TensorAllocationTest) {
   OrtValue* p_ml_value = frame.GetMutableNodeInputOrOutputMLValue(0);
   Tensor* p_tensor = p_ml_value ? p_ml_value->GetMutable<Tensor>() : nullptr;
   EXPECT_TRUE(p_tensor);
-  EXPECT_EQ(p_tensor->Shape(), shape);
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&p_tensor->Shape()), *reinterpret_cast<const std::vector<int64_t>*>(&shape));
   EXPECT_EQ(p_tensor->DataType(), DataTypeImpl::GetType<float>());
 
   //test share memory from tensor
@@ -106,7 +107,8 @@ TEST_F(ExecutionFrameTest, TensorAllocationTest) {
   const OrtValue* p_ml_value_const = frame.GetNodeInputOrOutputMLValue(1);
   auto tensor2 = p_ml_value_const ? &(p_ml_value_const->Get<Tensor>()) : nullptr;
   EXPECT_TRUE(tensor2);
-  EXPECT_EQ(tensor2->Shape(), shape2);
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&tensor2->Shape()), *reinterpret_cast<const std::vector<int64_t>*>(&shape2));
   EXPECT_EQ(tensor2->template Data<float>(), p_tensor->template Data<float>());
 }
 
@@ -155,7 +157,8 @@ TEST_F(ExecutionFrameTest, FeedInDataTest) {
   OrtValue* p_ml_value = frame.GetMutableNodeInputOrOutputMLValue(0);
   Tensor* p_tensor_arg_0 = p_ml_value ? p_ml_value->GetMutable<Tensor>() : nullptr;
   EXPECT_TRUE(p_tensor_arg_0);
-  EXPECT_EQ(p_tensor_arg_0->Shape(), shape);
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&p_tensor_arg_0->Shape()), *reinterpret_cast<const std::vector<int64_t>*>(&shape));
   EXPECT_EQ(p_tensor_arg_0->DataType(), DataTypeImpl::GetType<float>());
   EXPECT_EQ(p_tensor_arg_0->MutableData<float>(), value.GetMutable<Tensor>()->MutableData<float>());
 }
diff --git a/onnxruntime/test/framework/float_16_test.cc b/onnxruntime/test/framework/float_16_test.cc
index c141ea817c456..bb4ccc930afd0 100644
--- a/onnxruntime/test/framework/float_16_test.cc
+++ b/onnxruntime/test/framework/float_16_test.cc
@@ -122,7 +122,8 @@ void RunSession(InferenceSession& session_object,
   ASSERT_EQ(1, fetches.size());
   auto& rtensor = fetches.front().Get<Tensor>();
   TensorShape expected_shape(dims_y);
-  EXPECT_EQ(expected_shape, rtensor.Shape());
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&expected_shape), *reinterpret_cast<const std::vector<int64_t>*>(&rtensor.Shape()));
   const std::vector<MLFloat16> found(rtensor.template Data<MLFloat16>(), rtensor.template Data<MLFloat16>() + expected_shape.Size());
   ASSERT_EQ(found.size(), values_y.size());
   for (size_t i = 0; i < found.size(); i++)
diff --git a/onnxruntime/test/framework/inference_session_test.cc b/onnxruntime/test/framework/inference_session_test.cc
index ca5e2f8d7d36e..411bba33d987a 100644
--- a/onnxruntime/test/framework/inference_session_test.cc
+++ b/onnxruntime/test/framework/inference_session_test.cc
@@ -176,7 +176,8 @@ template <typename T = float>
 void VerifyOutputs(const Tensor& tensor, const std::vector<int64_t>& expected_dims,
                    const std::vector<T>& expected_values) {
   TensorShape expected_shape(expected_dims);
-  ASSERT_EQ(expected_shape, tensor.Shape());
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  ASSERT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&expected_shape), *reinterpret_cast<const std::vector<int64_t>*>(&tensor.Shape()));
   const std::vector<T> found(tensor.template Data<T>(),
                              tensor.template Data<T>() + expected_values.size());
   ASSERT_EQ(expected_values, found);
@@ -1312,7 +1313,8 @@ TEST(InferenceSessionTests, TestTruncatedSequence) {
   ASSERT_EQ(1, fetches.size());
   auto& rtensor = fetches.front().Get<Tensor>();
   TensorShape expected_shape(Y_dims);
-  ASSERT_EQ(expected_shape, rtensor.Shape());
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  ASSERT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&expected_shape), *reinterpret_cast<const std::vector<int64_t>*>(&rtensor.Shape()));
   for (size_t i = 0; i < Y_data.size(); ++i)
     EXPECT_NEAR(Y_data[i], rtensor.template Data<float>()[i], FLT_EPSILON);
 
@@ -1354,7 +1356,8 @@ TEST(InferenceSessionTests, TestTruncatedSequence) {
     std::vector<int64_t> truncated_output_dims = Y_dims;
     truncated_output_dims[0] = truncated_len;
     TensorShape truncated_shape(truncated_output_dims);
-    ASSERT_EQ(truncated_shape, truncated_rtensor.Shape());
+    //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+    ASSERT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&truncated_shape), *reinterpret_cast<const std::vector<int64_t>*>(&truncated_rtensor.Shape()));
     auto seq_output_stride = truncated_shape.SizeFromDimension(1);
     for (int i = 0; i < truncated_shape.Size(); ++i)
       EXPECT_NEAR(Y_data[i + seq_start * seq_output_stride], truncated_rtensor.template Data<float>()[i], FLT_EPSILON);
diff --git a/onnxruntime/test/framework/local_kernel_registry_test.cc b/onnxruntime/test/framework/local_kernel_registry_test.cc
index 09352e92b0bcd..9ef4d4de9bf25 100644
--- a/onnxruntime/test/framework/local_kernel_registry_test.cc
+++ b/onnxruntime/test/framework/local_kernel_registry_test.cc
@@ -214,7 +214,8 @@ void RunSession(InferenceSession& session_object,
   ASSERT_EQ(1, fetches.size());
   auto& rtensor = fetches.front().Get<Tensor>();
   TensorShape expected_shape(dims_y);
-  EXPECT_EQ(expected_shape, rtensor.Shape());
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&expected_shape), *reinterpret_cast<const std::vector<int64_t>*>(&rtensor.Shape()));
   const std::vector<float> found(rtensor.template Data<float>(), rtensor.template Data<float>() + expected_shape.Size());
   ASSERT_EQ(values_y, found);
 }
diff --git a/onnxruntime/test/framework/tensor_test.cc b/onnxruntime/test/framework/tensor_test.cc
index 1dea1a5e95993..d6b20a4b41e3a 100644
--- a/onnxruntime/test/framework/tensor_test.cc
+++ b/onnxruntime/test/framework/tensor_test.cc
@@ -21,7 +21,8 @@ void CPUTensorTest(std::vector<int64_t> dims, const int offset = 0) {
   EXPECT_TRUE(data);
   Tensor t(DataTypeImpl::GetType<T>(), shape, data, alloc->Info(), offset);
   auto tensor_shape = t.Shape();
-  EXPECT_EQ(shape, tensor_shape);
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&shape), *reinterpret_cast<const std::vector<int64_t>*>(&tensor_shape));
   EXPECT_EQ(t.DataType(), DataTypeImpl::GetType<T>());
   auto& location = t.Location();
   EXPECT_STREQ(location.name, CPU);
@@ -36,7 +37,8 @@ void CPUTensorTest(std::vector<int64_t> dims, const int offset = 0) {
   Tensor new_t(DataTypeImpl::GetType<T>(), shape, alloc, offset);
 
   tensor_shape = new_t.Shape();
-  EXPECT_EQ(shape, tensor_shape);
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&shape), *reinterpret_cast<const std::vector<int64_t>*>(&tensor_shape));
   EXPECT_EQ(new_t.DataType(), DataTypeImpl::GetType<T>());
   auto& new_location = new_t.Location();
   ASSERT_STREQ(new_location.name, CPU);
@@ -150,7 +152,8 @@ TEST(TensorTest, StringTensorTest) {
     Tensor t(DataTypeImpl::GetType<std::string>(), shape, alloc);
 
     auto& tensor_shape = t.Shape();
-    EXPECT_EQ(shape, tensor_shape);
+    //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+    EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&shape), *reinterpret_cast<const std::vector<int64_t>*>(&tensor_shape));
     EXPECT_EQ(t.DataType(), DataTypeImpl::GetType<std::string>());
     auto& location = t.Location();
     ASSERT_STREQ(location.name, CPU);
diff --git a/onnxruntime/test/optimizer/graph_transform_test.cc b/onnxruntime/test/optimizer/graph_transform_test.cc
index 46e3682c0c763..6bce5a239c0e1 100644
--- a/onnxruntime/test/optimizer/graph_transform_test.cc
+++ b/onnxruntime/test/optimizer/graph_transform_test.cc
@@ -678,7 +678,8 @@ TEST(GraphTransformationTests, FuseConvBnAddMulFloat16) {
   ASSERT_EQ(1, fetches.size());
   auto& rtensor = fetches.front().Get<Tensor>();
   TensorShape expected_shape(expected_dims_prod);
-  ASSERT_EQ(expected_shape, rtensor.Shape());
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  ASSERT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&expected_shape), *reinterpret_cast<const std::vector<int64_t>*>(&rtensor.Shape()));
   const std::vector<MLFloat16> found(rtensor.template Data<MLFloat16>(),
                                      rtensor.template Data<MLFloat16>() + expected_dims_prod.size());
   ASSERT_EQ(expected_values_prod, found);
diff --git a/onnxruntime/test/providers/cpu/controlflow/loop_test.cc b/onnxruntime/test/providers/cpu/controlflow/loop_test.cc
index e6813e2a07875..b6e1660364c23 100644
--- a/onnxruntime/test/providers/cpu/controlflow/loop_test.cc
+++ b/onnxruntime/test/providers/cpu/controlflow/loop_test.cc
@@ -620,7 +620,8 @@ TEST(Loop, SubgraphInputShadowsOuterScopeValue) {
 
   auto& b_out = fetches[0].Get<Tensor>();
   TensorShape expected_shape(scalar);
-  ASSERT_EQ(expected_shape, b_out.Shape());
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  ASSERT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&expected_shape), *reinterpret_cast<const std::vector<int64_t>*>(&b_out.Shape()));
   ASSERT_EQ(b_out.DataAsSpan<float>()[0], expected_value_b);
 
   auto user_defined_vals_out = fetches[1].Get<Tensor>().DataAsSpan<float>();

From b308e826a81040f0d817984a7c971bfff0f3e996 Mon Sep 17 00:00:00 2001
From: Pranav Sharma <prs@microsoft.com>
Date: Mon, 13 Jan 2020 20:26:16 -0800
Subject: [PATCH 55/75] Add support for int64_t for topk CPU. Fixes github
 issue #2806. (#2833)

---
 .../providers/cpu/cpu_execution_provider.cc   |  8 +--
 onnxruntime/core/providers/cpu/math/top_k.cc  | 52 ++++++++++++++-----
 2 files changed, 44 insertions(+), 16 deletions(-)

diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
index 496d96bba76dd..654173c54ea4d 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
@@ -416,7 +416,8 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain,
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, GatherND);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Range);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Unique);
-class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, TopK);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float, TopK);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int64_t, TopK);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int64_t_int64_t_int64_t, OneHot);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float_int64_t_int64_t, OneHot);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int64_t_string_int64_t, OneHot);
@@ -892,7 +893,7 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, int64_t,
                                                                   Where)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, uint8_t,
-                                                                  Where)>,                                                                  
+                                                                  Where)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
                                                                       Flatten)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9,
@@ -1076,7 +1077,8 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
       BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, GatherND)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Range)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Unique)>,
-      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int64_t, TopK)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
                                                                   int64_t_int64_t_int64_t, OneHot)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
diff --git a/onnxruntime/core/providers/cpu/math/top_k.cc b/onnxruntime/core/providers/cpu/math/top_k.cc
index 5e8264ef939ed..4eec811d85855 100644
--- a/onnxruntime/core/providers/cpu/math/top_k.cc
+++ b/onnxruntime/core/providers/cpu/math/top_k.cc
@@ -276,24 +276,33 @@ Status TopK<10, float>::Compute(OpKernelContext* p_op_kernel_context) const {
 }
 
 // Opset ver - 11
-template <>
-TopK<11, float>::TopK(const OpKernelInfo& op_kernel_info) : OpKernel(op_kernel_info) {
+
+static void TopkOpset11ConstructorCommon(const OpKernelInfo& op_kernel_info,
+                                         int& axis, bool& largest, bool& sorted) {
   int64_t axis_temp;
   ORT_ENFORCE(op_kernel_info.GetAttr<int64_t>("axis", &axis_temp).IsOK());
-  axis_ = gsl::narrow_cast<int>(axis_temp);
+  axis = gsl::narrow_cast<int>(axis_temp);
 
   int64_t largest_temp;
   ORT_ENFORCE(op_kernel_info.GetAttr<int64_t>("largest", &largest_temp).IsOK());
-  largest_ = largest_temp == 1 ? true : false;
+  largest = largest_temp == 1 ? true : false;
 
   int64_t sorted_temp;
   ORT_ENFORCE(op_kernel_info.GetAttr<int64_t>("sorted", &sorted_temp).IsOK());
-  sorted_ = sorted_temp == 1 ? true : false;
+  sorted = sorted_temp == 1 ? true : false;
 }
 
-// Opset ver - 11
 template <>
-Status TopK<11, float>::Compute(OpKernelContext* p_op_kernel_context) const {
+TopK<11, float>::TopK(const OpKernelInfo& op_kernel_info) : OpKernel(op_kernel_info) {
+  TopkOpset11ConstructorCommon(op_kernel_info, axis_, largest_, sorted_);
+}
+
+template <>
+TopK<11, int64_t>::TopK(const OpKernelInfo& op_kernel_info) : OpKernel(op_kernel_info) {
+  TopkOpset11ConstructorCommon(op_kernel_info, axis_, largest_, sorted_);
+}
+
+static Status ComputeImplOpset11(OpKernelContext* p_op_kernel_context, int axis, bool is_largest, bool is_sorted) {
   const auto* X = p_op_kernel_context->Input<Tensor>(0);
   const auto* Y = p_op_kernel_context->Input<Tensor>(1);
   if (X == nullptr || Y == nullptr) {
@@ -312,7 +321,18 @@ Status TopK<11, float>::Compute(OpKernelContext* p_op_kernel_context) const {
     return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL, "value of k must not be negative");
   }
 
-  return TopKImpl(p_op_kernel_context, X, axis_, gsl::narrow_cast<unsigned>(parsed_input_k), largest_, sorted_);
+  return TopKImpl(p_op_kernel_context, X, axis, gsl::narrow_cast<unsigned>(parsed_input_k), is_largest, is_sorted);
+}
+
+// Opset ver - 11
+template <>
+Status TopK<11, float>::Compute(OpKernelContext* p_op_kernel_context) const {
+  return ComputeImplOpset11(p_op_kernel_context, axis_, largest_, sorted_);
+}
+
+template <>
+Status TopK<11, int64_t>::Compute(OpKernelContext* p_op_kernel_context) const {
+  return ComputeImplOpset11(p_op_kernel_context, axis_, largest_, sorted_);
 }
 
 // Register necessary kernels
@@ -329,10 +349,16 @@ ONNX_CPU_OPERATOR_VERSIONED_KERNEL(TopK, 10, 10,
                                        .TypeConstraint("I", DataTypeImpl::GetTensorType<int64_t>()),
                                    TopK<10, float>);
 
-ONNX_CPU_OPERATOR_KERNEL(TopK, 11,
-                         KernelDefBuilder()
-                             .TypeConstraint("T", DataTypeImpl::GetTensorType<float>())
-                             .TypeConstraint("I", DataTypeImpl::GetTensorType<int64_t>()),
-                         TopK<11, float>);
+#define REGISTER_TOPK_TYPED_KERNEL(OPSET, TYPE)                                                    \
+  ONNX_CPU_OPERATOR_TYPED_KERNEL(TopK,                                                             \
+                                 OPSET,                                                            \
+                                 TYPE,                                                             \
+                                 KernelDefBuilder()                                                \
+                                     .TypeConstraint("T", DataTypeImpl::GetTensorType<TYPE>())     \
+                                     .TypeConstraint("I", DataTypeImpl::GetTensorType<int64_t>()), \
+                                 TopK<OPSET, TYPE>);
+
+REGISTER_TOPK_TYPED_KERNEL(11, float);
+REGISTER_TOPK_TYPED_KERNEL(11, int64_t);
 
 }  // namespace onnxruntime

From 98cb41aa039b08c5d2451dd92124ab9a480f4099 Mon Sep 17 00:00:00 2001
From: Scott McKay <skottmckay@gmail.com>
Date: Tue, 14 Jan 2020 18:14:55 +1000
Subject: [PATCH 56/75] Ignore allocator type in ExecutionProviders allocator
 map. Make default initialization of OrtMemoryInfo more clearly invalid.
 (#2768)

* Remove allocator type from the key comparison in ExecutionProviders.
Remove usage of DummyArena as it's no longer necessary.

* Fix x86 tests where arena allocator is disabled.
Make initialization of OrtMemoryInfo clearer by adding Invalid enum value.

* Make OrtValueNameIdxMap::MaxIdx more intuitive.
---
 .../onnxruntime/core/framework/allocator.h    |  40 +++--
 .../core/session/onnxruntime_c_api.h          | 151 +++++++++---------
 onnxruntime/core/framework/allocator.cc       |   2 +-
 .../core/framework/execution_providers.h      |  19 ++-
 .../core/framework/ort_value_name_idx_map.h   |   8 +-
 .../framework/sequential_execution_plan.h     |   2 +-
 .../framework/session_state_initializer.cc    |   4 +-
 .../providers/cpu/cpu_execution_provider.h    |   9 +-
 onnxruntime/test/framework/allocator_test.cc  |   8 +-
 .../framework/cuda/allocator_cuda_test.cc     |  18 ++-
 .../test/framework/inference_session_test.cc  |   7 +-
 onnxruntime/test/framework/tensor_test.cc     |   8 +-
 12 files changed, 155 insertions(+), 121 deletions(-)

diff --git a/include/onnxruntime/core/framework/allocator.h b/include/onnxruntime/core/framework/allocator.h
index f6a31494768c8..1fdcdfc02e313 100644
--- a/include/onnxruntime/core/framework/allocator.h
+++ b/include/onnxruntime/core/framework/allocator.h
@@ -85,26 +85,27 @@ struct OrtMemoryInfo {
 
   // use string for name, so we could have customized allocator in execution provider.
   const char* name;
-  int id;
-  OrtMemType mem_type;
-  OrtAllocatorType type;
+  int id = -1;
+  OrtMemType mem_type = OrtMemTypeDefault;
+  OrtAllocatorType alloc_type = Invalid;
   OrtDevice device;
 
-  constexpr OrtMemoryInfo(const char* name_, OrtAllocatorType type_, OrtDevice device_ = OrtDevice(), int id_ = 0, OrtMemType mem_type_ = OrtMemTypeDefault)
+  constexpr OrtMemoryInfo(const char* name_, OrtAllocatorType type_, OrtDevice device_ = OrtDevice(), int id_ = 0,
+                          OrtMemType mem_type_ = OrtMemTypeDefault)
 #if (defined(__GNUC__) || defined(__clang__))
       __attribute__((nonnull))
 #endif
       : name(name_),
         id(id_),
         mem_type(mem_type_),
-        type(type_),
+        alloc_type(type_),
         device(device_) {
   }
 
   // To make OrtMemoryInfo become a valid key in std map
-  inline bool operator<(const OrtMemoryInfo& other) const {
-    if (type != other.type)
-      return type < other.type;
+  bool operator<(const OrtMemoryInfo& other) const {
+    if (alloc_type != other.alloc_type)
+      return alloc_type < other.alloc_type;
     if (mem_type != other.mem_type)
       return mem_type < other.mem_type;
     if (id != other.id)
@@ -113,20 +114,22 @@ struct OrtMemoryInfo {
     return strcmp(name, other.name) < 0;
   }
 
-  inline std::string ToString() const {
+  std::string ToString() const {
     std::ostringstream ostr;
     ostr << "OrtMemoryInfo: ["
          << " name:" << name
          << " id:" << id
          << " mem_type:" << mem_type
-         << " type:" << type
+         << " alloc_type:" << alloc_type
          << "]";
     return ostr.str();
   }
 };
 
 inline bool operator==(const OrtMemoryInfo& left, const OrtMemoryInfo& other) {
-  return left.mem_type == other.mem_type && left.type == other.type && left.id == other.id &&
+  return left.mem_type == other.mem_type &&
+         left.alloc_type == other.alloc_type &&
+         left.id == other.id &&
          strcmp(left.name, other.name) == 0;
 }
 
@@ -213,9 +216,11 @@ class IAllocator {
     if (!std::is_void<T>::value) {
       // sizeof(void) isn't valid, but the compiler isn't smart enough to ignore that this line isn't
       // reachable if T is void. use std::conditional to 'use' void* in the sizeof call
-      if (!CalcMemSizeForArray(count_or_bytes, sizeof(typename std::conditional<std::is_void<T>::value, void*, T>::type),
+      if (!CalcMemSizeForArray(count_or_bytes,
+                               sizeof(typename std::conditional<std::is_void<T>::value, void*, T>::type),
                                &alloc_size)) return nullptr;
     }
+
     return IAllocatorUniquePtr<T>{
         static_cast<T*>(allocator->Alloc(alloc_size)),  // allocate
         [=](T* ptr) { allocator->Free(ptr); }};         // capture IAllocator so it's always valid, and use as deleter
@@ -224,22 +229,26 @@ class IAllocator {
 
 template <size_t alignment>
 bool IAllocator::CalcMemSizeForArrayWithAlignment(size_t nmemb, size_t size, size_t* out) noexcept {
-  static constexpr size_t max_allowed = (static_cast<size_t>(1) << (static_cast<size_t>(std::numeric_limits<size_t>::digits >> 1))) - alignment;
+  static constexpr size_t max_allowed = (size_t(1) << (size_t(std::numeric_limits<size_t>::digits >> 1))) - alignment;
   static constexpr size_t max_size = std::numeric_limits<size_t>::max() - alignment;
   static constexpr size_t alignment_mask = alignment - 1;
+
   //Indeed, we only need to check if max_size / nmemb < size
   //max_allowed is for avoiding unnecessary DIV.
   if (nmemb >= max_allowed && max_size / nmemb < size) {
     return false;
   }
+
   if (size >= max_allowed &&
       nmemb > 0 && max_size / nmemb < size) {
     return false;
   }
+
   if (alignment == 0)
     *out = size * nmemb;
   else
     *out = (size * nmemb + alignment_mask) & ~static_cast<size_t>(alignment_mask);
+
   return true;
 }
 
@@ -297,11 +306,10 @@ class MiMallocAllocator : public IDeviceAllocator {
 
 #endif
 
-
 #ifdef USE_MIMALLOC
-  using TAllocator = MiMallocAllocator;
+using TAllocator = MiMallocAllocator;
 #else
-  using TAllocator = CPUAllocator;
+using TAllocator = CPUAllocator;
 #endif
 
 using AllocatorPtr = std::shared_ptr<IAllocator>;
diff --git a/include/onnxruntime/core/session/onnxruntime_c_api.h b/include/onnxruntime/core/session/onnxruntime_c_api.h
index ed07c3bbb2d87..247d204f9b0de 100644
--- a/include/onnxruntime/core/session/onnxruntime_c_api.h
+++ b/include/onnxruntime/core/session/onnxruntime_c_api.h
@@ -84,12 +84,12 @@ typedef enum ONNXTensorElementDataType {
   ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING,  // maps to c++ type std::string
   ONNX_TENSOR_ELEMENT_DATA_TYPE_BOOL,
   ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT16,
-  ONNX_TENSOR_ELEMENT_DATA_TYPE_DOUBLE,          // maps to c type double
-  ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT32,          // maps to c type uint32_t
-  ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT64,          // maps to c type uint64_t
-  ONNX_TENSOR_ELEMENT_DATA_TYPE_COMPLEX64,       // complex with float32 real and imaginary components
-  ONNX_TENSOR_ELEMENT_DATA_TYPE_COMPLEX128,      // complex with float64 real and imaginary components
-  ONNX_TENSOR_ELEMENT_DATA_TYPE_BFLOAT16        // Non-IEEE floating-point format based on IEEE754 single-precision
+  ONNX_TENSOR_ELEMENT_DATA_TYPE_DOUBLE,      // maps to c type double
+  ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT32,      // maps to c type uint32_t
+  ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT64,      // maps to c type uint64_t
+  ONNX_TENSOR_ELEMENT_DATA_TYPE_COMPLEX64,   // complex with float32 real and imaginary components
+  ONNX_TENSOR_ELEMENT_DATA_TYPE_COMPLEX128,  // complex with float64 real and imaginary components
+  ONNX_TENSOR_ELEMENT_DATA_TYPE_BFLOAT16     // Non-IEEE floating-point format based on IEEE754 single-precision
 } ONNXTensorElementDataType;
 
 // Synced with onnx TypeProto oneof
@@ -193,6 +193,7 @@ struct OrtCustomOp;
 typedef struct OrtCustomOp OrtCustomOp;
 
 typedef enum OrtAllocatorType {
+  Invalid = -1,
   OrtDeviceAllocator = 0,
   OrtArenaAllocator = 1
 } OrtAllocatorType;
@@ -234,14 +235,14 @@ struct OrtApi {
   const char*(ORT_API_CALL* GetErrorMessage)(_In_ const OrtStatus* status)NO_EXCEPTION ORT_ALL_ARGS_NONNULL;
 
   /**
-	 * \param out Should be freed by `OrtReleaseEnv` after use
-	 */
+     * \param out Should be freed by `OrtReleaseEnv` after use
+     */
   OrtStatus*(ORT_API_CALL* CreateEnv)(OrtLoggingLevel default_logging_level, _In_ const char* logid, _Outptr_ OrtEnv** out)
       NO_EXCEPTION ORT_ALL_ARGS_NONNULL;
 
   /**
-	 * \param out Should be freed by `OrtReleaseEnv` after use
-	 */
+   * \param out Should be freed by `OrtReleaseEnv` after use
+   */
   OrtStatus*(ORT_API_CALL* CreateEnvWithCustomLogger)(OrtLoggingFunction logging_function,
                                                       _In_opt_ void* logger_param, OrtLoggingLevel default_warning_level,
                                                       _In_ const char* logid,
@@ -268,8 +269,8 @@ struct OrtApi {
                                 _In_ const char* const* output_names, size_t output_names_len, _Outptr_ OrtValue** output)NO_EXCEPTION;
 
   /**
-	* \return A pointer of the newly created object. The pointer should be freed by OrtReleaseSessionOptions after use
-	*/
+    * \return A pointer of the newly created object. The pointer should be freed by OrtReleaseSessionOptions after use
+    */
   OrtStatus*(ORT_API_CALL* CreateSessionOptions)(_Outptr_ OrtSessionOptions** options)NO_EXCEPTION;
 
   // Set filepath to save optimized model after graph level transformations.
@@ -325,36 +326,36 @@ struct OrtApi {
   OrtStatus*(ORT_API_CALL* CreateCustomOpDomain)(_In_ const char* domain, _Outptr_ OrtCustomOpDomain** out)NO_EXCEPTION;
 
   /*
-	 * Add custom ops to the OrtCustomOpDomain
-	 *  Note: The OrtCustomOp* pointer must remain valid until the OrtCustomOpDomain using it is released
-	*/
+     * Add custom ops to the OrtCustomOpDomain
+     *  Note: The OrtCustomOp* pointer must remain valid until the OrtCustomOpDomain using it is released
+    */
   OrtStatus*(ORT_API_CALL* CustomOpDomain_Add)(_Inout_ OrtCustomOpDomain* custom_op_domain, _In_ OrtCustomOp* op)NO_EXCEPTION;
 
   /*
-	 * Add a custom op domain to the OrtSessionOptions
-	 *  Note: The OrtCustomOpDomain* must not be deleted until the sessions using it are released
-	*/
+     * Add a custom op domain to the OrtSessionOptions
+     *  Note: The OrtCustomOpDomain* must not be deleted until the sessions using it are released
+    */
   OrtStatus*(ORT_API_CALL* AddCustomOpDomain)(_Inout_ OrtSessionOptions* options, _In_ OrtCustomOpDomain* custom_op_domain)NO_EXCEPTION;
 
   /*
-	 * Loads a DLL named 'library_path' and looks for this entry point:
-	 *		OrtStatus* RegisterCustomOps(OrtSessionOptions * options, const OrtApiBase* api);
-	 * It then passes in the provided session options to this function along with the api base.
-	 * The handle to the loaded library is returned in library_handle. It can be freed by the caller after all sessions using the passed in
-	 * session options are destroyed, or if an error occurs and it is non null.
+     * Loads a DLL named 'library_path' and looks for this entry point:
+     *		OrtStatus* RegisterCustomOps(OrtSessionOptions * options, const OrtApiBase* api);
+     * It then passes in the provided session options to this function along with the api base.
+     * The handle to the loaded library is returned in library_handle. It can be freed by the caller after all sessions using the passed in
+     * session options are destroyed, or if an error occurs and it is non null.
   */
   OrtStatus*(ORT_API_CALL* RegisterCustomOpsLibrary)(_Inout_ OrtSessionOptions* options, _In_ const char* library_path, void** library_handle)NO_EXCEPTION;
 
   /**
-	* To use additional providers, you must build ORT with the extra providers enabled. Then call one of these
-	* functions to enable them in the session:
-	*   OrtSessionOptionsAppendExecutionProvider_CPU
-	*   OrtSessionOptionsAppendExecutionProvider_CUDA
-	*   OrtSessionOptionsAppendExecutionProvider_<remaining providers...>
-	* The order they care called indicates the preference order as well. In other words call this method
-	* on your most preferred execution provider first followed by the less preferred ones.
-	* If none are called Ort will use its internal CPU execution provider.
-	*/
+    * To use additional providers, you must build ORT with the extra providers enabled. Then call one of these
+    * functions to enable them in the session:
+    *   OrtSessionOptionsAppendExecutionProvider_CPU
+    *   OrtSessionOptionsAppendExecutionProvider_CUDA
+    *   OrtSessionOptionsAppendExecutionProvider_<remaining providers...>
+    * The order they care called indicates the preference order as well. In other words call this method
+    * on your most preferred execution provider first followed by the less preferred ones.
+    * If none are called Ort will use its internal CPU execution provider.
+    */
 
   OrtStatus*(ORT_API_CALL* SessionGetInputCount)(_In_ const OrtSession* sess, _Out_ size_t* out)NO_EXCEPTION;
   OrtStatus*(ORT_API_CALL* SessionGetOutputCount)(_In_ const OrtSession* sess, _Out_ size_t* out)NO_EXCEPTION;
@@ -432,39 +433,39 @@ struct OrtApi {
   OrtStatus*(ORT_API_CALL* GetTensorMutableData)(_Inout_ OrtValue* value, _Outptr_ void** out)NO_EXCEPTION;
 
   /**
-	 * \param value A tensor created from CreateTensor... function.
-	 * \param s each A string array. Each string in this array must be null terminated.
-	 * \param s_len length of s
-	 */
+     * \param value A tensor created from OrtCreateTensor... function.
+     * \param s each A string array. Each string in this array must be null terminated.
+     * \param s_len length of s
+     */
   OrtStatus*(ORT_API_CALL* FillStringTensor)(_Inout_ OrtValue* value, _In_ const char* const* s, size_t s_len)NO_EXCEPTION;
 
   /**
-	 * \param value A tensor created from CreateTensor... function.
-	 * \param len total data length, not including the trailing '\0' chars.
-	 */
+     * \param value A tensor created from OrtCreateTensor... function.
+     * \param len total data length, not including the trailing '\0' chars.
+     */
   OrtStatus*(ORT_API_CALL* GetStringTensorDataLength)(_In_ const OrtValue* value, _Out_ size_t* len)NO_EXCEPTION;
 
   /**
-	 * \param s string contents. Each string is NOT null-terminated.
-	 * \param value A tensor created from CreateTensor... function.
-	 * \param s_len total data length, get it from GetStringTensorDataLength
-	 */
+     * \param s string contents. Each string is NOT null-terminated.
+     * \param value A tensor created from OrtCreateTensor... function.
+     * \param s_len total data length, get it from OrtGetStringTensorDataLength
+     */
   OrtStatus*(ORT_API_CALL* GetStringTensorContent)(_In_ const OrtValue* value, _Out_ void* s, size_t s_len,
                                                    _Out_ size_t* offsets, size_t offsets_len)NO_EXCEPTION;
 
   /**
-	 * Don't free the 'out' value
-	 */
+     * Don't free the 'out' value
+     */
   OrtStatus*(ORT_API_CALL* CastTypeInfoToTensorInfo)(_In_ const OrtTypeInfo*, _Out_ const OrtTensorTypeAndShapeInfo** out)NO_EXCEPTION;
 
   /**
-	 * Return OnnxType from OrtTypeInfo
-	 */
+     * Return OnnxType from OrtTypeInfo
+     */
   OrtStatus*(ORT_API_CALL* GetOnnxTypeFromTypeInfo)(_In_ const OrtTypeInfo*, _Out_ enum ONNXType* out)NO_EXCEPTION;
 
   /**
-	 * The 'out' value should be released by calling OrtReleaseTensorTypeAndShapeInfo
-	 */
+     * The 'out' value should be released by calling OrtReleaseTensorTypeAndShapeInfo
+     */
   OrtStatus*(ORT_API_CALL* CreateTensorTypeAndShapeInfo)(_Outptr_ OrtTensorTypeAndShapeInfo** out)NO_EXCEPTION;
 
   OrtStatus*(ORT_API_CALL* SetTensorElementType)(_Inout_ OrtTensorTypeAndShapeInfo*, enum ONNXTensorElementDataType type)NO_EXCEPTION;
@@ -592,36 +593,36 @@ struct OrtApi {
                                         _Outptr_ OrtValue** out)NO_EXCEPTION;
 
   /**
-	 * Construct OrtValue that contains a value of non-standard type created for
-	 * experiments or while awaiting standardization. OrtValue in this case would contain
-	 * an internal representation of the Opaque type. Opaque types are distinguished between
-	 * each other by two strings 1) domain and 2) type name. The combination of the two
-	 * must be unique, so the type representation is properly identified internally. The combination
-	 * must be properly registered from within ORT at both compile/run time or by another API.
-	 *
-	 * To construct the OrtValue pass domain and type names, also a pointer to a data container
-	 * the type of which must be know to both ORT and the client program. That data container may or may
-	 * not match the internal representation of the Opaque type. The sizeof(data_container) is passed for
-	 * verification purposes.
-	 *
-	 * \domain_name - domain name for the Opaque type, null terminated.
-	 * \type_name   - type name for the Opaque type, null terminated.
-	 * \data_contianer - data to populate OrtValue
-	 * \data_container_size - sizeof() of the data container. Must match the sizeof() of the expected
-	 *                    data_container size internally.
-	 */
+     * Construct OrtValue that contains a value of non-standard type created for
+     * experiments or while awaiting standardization. OrtValue in this case would contain
+     * an internal representation of the Opaque type. Opaque types are distinguished between
+     * each other by two strings 1) domain and 2) type name. The combination of the two
+     * must be unique, so the type representation is properly identified internally. The combination
+     * must be properly registered from within ORT at both compile/run time or by another API.
+     *
+     * To construct the OrtValue pass domain and type names, also a pointer to a data container
+     * the type of which must be know to both ORT and the client program. That data container may or may
+     * not match the internal representation of the Opaque type. The sizeof(data_container) is passed for
+     * verification purposes.
+     *
+     * \domain_name - domain name for the Opaque type, null terminated.
+     * \type_name   - type name for the Opaque type, null terminated.
+     * \data_contianer - data to populate OrtValue
+     * \data_container_size - sizeof() of the data container. Must match the sizeof() of the expected
+     *                    data_container size internally.
+     */
   OrtStatus*(ORT_API_CALL* CreateOpaqueValue)(_In_ const char* domain_name, _In_ const char* type_name,
                                               _In_ const void* data_container, size_t data_container_size, _Outptr_ OrtValue** out)NO_EXCEPTION;
 
   /**
-	 * Fetch data from an OrtValue that contains a value of non-standard type created for
-	 * experiments or while awaiting standardization.
-	 * \domain_name - domain name for the Opaque type, null terminated.
-	 * \type_name   - type name for the Opaque type, null terminated.
-	 * \data_contianer - data to populate OrtValue
-	 * \data_container_size - sizeof() of the data container. Must match the sizeof() of the expected
-	 *                    data_container size internally.
-	 */
+     * Fetch data from an OrtValue that contains a value of non-standard type created for
+     * experiments or while awaiting standardization.
+     * \domain_name - domain name for the Opaque type, null terminated.
+     * \type_name   - type name for the Opaque type, null terminated.
+     * \data_contianer - data to populate OrtValue
+     * \data_container_size - sizeof() of the data container. Must match the sizeof() of the expected
+     *                    data_container size internally.
+     */
 
   OrtStatus*(ORT_API_CALL* GetOpaqueValue)(_In_ const char* domain_name, _In_ const char* type_name,
                                            _In_ const OrtValue* in, _Out_ void* data_container, size_t data_container_size)NO_EXCEPTION;
diff --git a/onnxruntime/core/framework/allocator.cc b/onnxruntime/core/framework/allocator.cc
index f58020df3f426..d3a567c4a207d 100644
--- a/onnxruntime/core/framework/allocator.cc
+++ b/onnxruntime/core/framework/allocator.cc
@@ -71,7 +71,7 @@ ORT_API_STATUS_IMPL(OrtApis::MemoryInfoGetMemType, _In_ const OrtMemoryInfo* ptr
 }
 
 ORT_API_STATUS_IMPL(OrtApis::MemoryInfoGetType, _In_ const OrtMemoryInfo* ptr, _Out_ OrtAllocatorType* out) {
-  *out = ptr->type;
+  *out = ptr->alloc_type;
   return nullptr;
 }
 
diff --git a/onnxruntime/core/framework/execution_providers.h b/onnxruntime/core/framework/execution_providers.h
index ec7e2fedc9e2a..b17c3a61c046d 100644
--- a/onnxruntime/core/framework/execution_providers.h
+++ b/onnxruntime/core/framework/execution_providers.h
@@ -107,8 +107,25 @@ class ExecutionProviders {
 
   // maps for fast lookup of an index into exec_providers_
   std::unordered_map<std::string, size_t> provider_idx_map_;
+
+  // currently the allocator type is an implementation detail and we don't make any  behavioral choices based on it,
+  // so exclude it from the key comparison for allocator_idx_map_.
+  // we also don't expect to have two allocators with the same name, one using an arena and one not.
+  struct OrtMemoryInfoLessThanIgnoreAllocType {
+    bool operator()(const OrtMemoryInfo& lhs, const OrtMemoryInfo& rhs) const {
+      /*if (lhs.alloc_type != rhs.alloc_type)
+        return lhs.alloc_type < rhs.alloc_type;*/
+      if (lhs.mem_type != rhs.mem_type)
+        return lhs.mem_type < rhs.mem_type;
+      if (lhs.id != rhs.id)
+        return lhs.id < rhs.id;
+
+      return strcmp(lhs.name, rhs.name) < 0;
+    }
+  };
+
   // using std::map as OrtMemoryInfo would need a custom hash function to be used with unordered_map,
   // and as this isn't performance critical it's not worth the maintenance overhead of adding one.
-  std::map<OrtMemoryInfo, size_t> allocator_idx_map_;
+  std::map<OrtMemoryInfo, size_t, OrtMemoryInfoLessThanIgnoreAllocType> allocator_idx_map_;
 };
 }  // namespace onnxruntime
diff --git a/onnxruntime/core/framework/ort_value_name_idx_map.h b/onnxruntime/core/framework/ort_value_name_idx_map.h
index dc906642d67e9..74e41c79a3f40 100644
--- a/onnxruntime/core/framework/ort_value_name_idx_map.h
+++ b/onnxruntime/core/framework/ort_value_name_idx_map.h
@@ -22,8 +22,7 @@ class OrtValueNameIdxMap {
   int Add(const std::string& name) {
     auto it = map_.find(name);
     if (it == map_.end()) {
-      int idx;
-      idx = ort_value_max_idx_++;
+      int idx = next_idx_++;
       map_.insert(it, {name, idx});
       return idx;
     }
@@ -43,7 +42,7 @@ class OrtValueNameIdxMap {
   }
 
   size_t Size() const { return map_.size(); };
-  int MaxIdx() const { return ort_value_max_idx_; }
+  int MaxIdx() const { return next_idx_ - 1; }
 
   const_iterator begin() const noexcept { return map_.cbegin(); }
   const_iterator end() const noexcept { return map_.cend(); }
@@ -51,8 +50,7 @@ class OrtValueNameIdxMap {
  private:
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(OrtValueNameIdxMap);
 
-  int ort_value_max_idx_ = 0;
+  int next_idx_ = 0;
   std::unordered_map<std::string, int> map_;
 };
-using OrtValueNameIdxMap = OrtValueNameIdxMap;
 }  // namespace onnxruntime
diff --git a/onnxruntime/core/framework/sequential_execution_plan.h b/onnxruntime/core/framework/sequential_execution_plan.h
index 3e3dca68020cb..924b62d395432 100644
--- a/onnxruntime/core/framework/sequential_execution_plan.h
+++ b/onnxruntime/core/framework/sequential_execution_plan.h
@@ -33,7 +33,7 @@ struct AllocPlanPerValue {
   bool create_fence_if_async{false};
 
  public:
-  AllocPlanPerValue() : location(CPU, OrtArenaAllocator) {}
+  AllocPlanPerValue() : location(CPU, Invalid) {}
 };
 
 // SequentialExecutionPlan: This is the data that is produced by a static
diff --git a/onnxruntime/core/framework/session_state_initializer.cc b/onnxruntime/core/framework/session_state_initializer.cc
index 3dd6ad72691fa..a80d6b6d47c85 100644
--- a/onnxruntime/core/framework/session_state_initializer.cc
+++ b/onnxruntime/core/framework/session_state_initializer.cc
@@ -150,7 +150,7 @@ static common::Status DeserializeTensorProto(const Env& env, const std::basic_st
   const Tensor& p_deserialize_tensor = tmp_ort_value.Get<Tensor>();
 
   p_tensor = onnxruntime::make_unique<Tensor>(p_deserialize_tensor.DataType(), p_deserialize_tensor.Shape(), m.GetBuffer(),
-                                      m.GetAllocInfo());
+                                              m.GetAllocInfo());
   // TODO: does this function work for string tensor?
   Status copy_status = data_transfer_mgr.CopyTensor(p_deserialize_tensor, *p_tensor);
   if (d.f) d.f(d.param);
@@ -177,7 +177,7 @@ common::Status SaveInitializedTensors(const Env& env, const std::basic_string<PA
                                       const T& save_tensor_func, const logging::Logger& logger,
                                       const DataTransferManager& data_transfer_mgr) {
   LOGS(logger, INFO) << "Saving initialized tensors.";
-  ORT_ENFORCE(ort_value_name_idx_map.MaxIdx() > 0, "OrtValue indexes should have been populated.");
+  ORT_ENFORCE(ort_value_name_idx_map.MaxIdx() > -1, "OrtValue indexes should have been populated.");
 
   //1. first plan the memory
   const onnxruntime::InitializedTensorSet& initialized_tensor_set = graph.GetAllInitializedTensors();
diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.h b/onnxruntime/core/providers/cpu/cpu_execution_provider.h
index c7d36d1f9bfa8..79099e65ba844 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.h
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.h
@@ -38,9 +38,7 @@ class CPUExecutionProvider : public IExecutionProvider {
 
     ORT_UNUSED_PARAMETER(info);
     //JEMalloc already has memory pool, so just use device allocator.
-    InsertAllocator(
-        std::shared_ptr<IArenaAllocator>(
-            onnxruntime::make_unique<DummyArena>(device_info.factory(0))));
+    InsertAllocator(device_info.factory(0));
 #else
 //Disable Arena allocator for x86_32 build because it may run into infinite loop when integer overflow happens
 #if defined(__amd64__) || defined(_M_AMD64)
@@ -50,10 +48,7 @@ class CPUExecutionProvider : public IExecutionProvider {
 #else
     ORT_UNUSED_PARAMETER(info);
 #endif
-      InsertAllocator(
-          std::shared_ptr<IArenaAllocator>(
-              onnxruntime::make_unique<DummyArena>(device_info.factory(0))));
-
+      InsertAllocator(device_info.factory(0));
 #endif
   }
 
diff --git a/onnxruntime/test/framework/allocator_test.cc b/onnxruntime/test/framework/allocator_test.cc
index 6fe5faf810ae9..9a5ded2146b7b 100644
--- a/onnxruntime/test/framework/allocator_test.cc
+++ b/onnxruntime/test/framework/allocator_test.cc
@@ -13,7 +13,13 @@ TEST(AllocatorTest, CPUAllocatorTest) {
 
   ASSERT_STREQ(cpu_arena->Info().name, CPU);
   EXPECT_EQ(cpu_arena->Info().id, 0);
-  EXPECT_EQ(cpu_arena->Info().type, OrtAllocatorType::OrtArenaAllocator);
+
+  // arena is disabled for CPUExecutionProvider on x86 and JEMalloc
+#if (defined(__amd64__) || defined(_M_AMD64)) && !defined(USE_JEMALLOC)
+  EXPECT_EQ(cpu_arena->Info().alloc_type, OrtAllocatorType::OrtArenaAllocator);
+#else
+  EXPECT_EQ(cpu_arena->Info().alloc_type, OrtAllocatorType::OrtDeviceAllocator);
+#endif
 
   size_t size = 1024;
   auto bytes = cpu_arena->Alloc(size);
diff --git a/onnxruntime/test/framework/cuda/allocator_cuda_test.cc b/onnxruntime/test/framework/cuda/allocator_cuda_test.cc
index 4bcca7ee7bd66..b0388b457a156 100644
--- a/onnxruntime/test/framework/cuda/allocator_cuda_test.cc
+++ b/onnxruntime/test/framework/cuda/allocator_cuda_test.cc
@@ -11,8 +11,10 @@ namespace onnxruntime {
 namespace test {
 TEST(AllocatorTest, CUDAAllocatorTest) {
   int cuda_device_id = 0;
-  DeviceAllocatorRegistrationInfo default_memory_info({OrtMemTypeDefault,
-                                                          [](int id) { return onnxruntime::make_unique<CUDAAllocator>(id, CUDA); }, std::numeric_limits<size_t>::max()});
+  DeviceAllocatorRegistrationInfo default_memory_info(
+      {OrtMemTypeDefault,
+       [](int id) { return onnxruntime::make_unique<CUDAAllocator>(id, CUDA); },
+       std::numeric_limits<size_t>::max()});
 
   auto cuda_arena = CreateAllocator(default_memory_info, cuda_device_id);
 
@@ -21,21 +23,23 @@ TEST(AllocatorTest, CUDAAllocatorTest) {
   EXPECT_STREQ(cuda_arena->Info().name, CUDA);
   EXPECT_EQ(cuda_arena->Info().id, cuda_device_id);
   EXPECT_EQ(cuda_arena->Info().mem_type, OrtMemTypeDefault);
-  EXPECT_EQ(cuda_arena->Info().type, OrtArenaAllocator);
+  EXPECT_EQ(cuda_arena->Info().alloc_type, OrtArenaAllocator);
 
   //test cuda allocation
   auto cuda_addr = cuda_arena->Alloc(size);
   EXPECT_TRUE(cuda_addr);
 
-  DeviceAllocatorRegistrationInfo pinned_memory_info({OrtMemTypeCPUOutput,
-                                                         [](int) { return onnxruntime::make_unique<CUDAPinnedAllocator>(0, CUDA_PINNED); }, std::numeric_limits<size_t>::max()});
+  DeviceAllocatorRegistrationInfo pinned_memory_info(
+      {OrtMemTypeCPUOutput,
+       [](int) { return onnxruntime::make_unique<CUDAPinnedAllocator>(0, CUDA_PINNED); },
+       std::numeric_limits<size_t>::max()});
 
   auto pinned_allocator = CreateAllocator(pinned_memory_info);
 
   EXPECT_STREQ(pinned_allocator->Info().name, CUDA_PINNED);
   EXPECT_EQ(pinned_allocator->Info().id, 0);
   EXPECT_EQ(pinned_allocator->Info().mem_type, OrtMemTypeCPUOutput);
-  EXPECT_EQ(pinned_allocator->Info().type, OrtArenaAllocator);
+  EXPECT_EQ(pinned_allocator->Info().alloc_type, OrtArenaAllocator);
 
   //test pinned allocation
   auto pinned_addr = pinned_allocator->Alloc(size);
@@ -45,7 +49,7 @@ TEST(AllocatorTest, CUDAAllocatorTest) {
   EXPECT_STREQ(cpu_arena->Info().name, CPU);
   EXPECT_EQ(cpu_arena->Info().id, 0);
   EXPECT_EQ(cpu_arena->Info().mem_type, OrtMemTypeDefault);
-  EXPECT_EQ(cpu_arena->Info().type, OrtArenaAllocator);
+  EXPECT_EQ(cpu_arena->Info().alloc_type, OrtArenaAllocator);
 
   auto cpu_addr_a = cpu_arena->Alloc(size);
   EXPECT_TRUE(cpu_addr_a);
diff --git a/onnxruntime/test/framework/inference_session_test.cc b/onnxruntime/test/framework/inference_session_test.cc
index 411bba33d987a..2063852cdea84 100644
--- a/onnxruntime/test/framework/inference_session_test.cc
+++ b/onnxruntime/test/framework/inference_session_test.cc
@@ -86,8 +86,7 @@ class FuseExecutionProvider : public IExecutionProvider {
     DeviceAllocatorRegistrationInfo device_info({OrtMemTypeDefault,
                                                  [](int) { return onnxruntime::make_unique<CPUAllocator>(); },
                                                  std::numeric_limits<size_t>::max()});
-    InsertAllocator(std::shared_ptr<IArenaAllocator>(
-        onnxruntime::make_unique<DummyArena>(device_info.factory(0))));
+    InsertAllocator(device_info.factory(0));
   }
 
   std::vector<std::unique_ptr<ComputeCapability>>
@@ -776,8 +775,8 @@ static void TestBindHelper(const std::string& log_str,
   std::string s1;
   p_model->ToProto().SerializeToString(&s1);
   std::stringstream sstr(s1);
-  ASSERT_TRUE(session_object.Load(sstr).IsOK());
-  ASSERT_TRUE(session_object.Initialize().IsOK());
+  ASSERT_STATUS_OK(session_object.Load(sstr));
+  ASSERT_STATUS_OK(session_object.Initialize());
 
   RunOptions run_options;
   run_options.run_log_verbosity_level = so.session_log_verbosity_level;
diff --git a/onnxruntime/test/framework/tensor_test.cc b/onnxruntime/test/framework/tensor_test.cc
index d6b20a4b41e3a..6251b0259e719 100644
--- a/onnxruntime/test/framework/tensor_test.cc
+++ b/onnxruntime/test/framework/tensor_test.cc
@@ -136,7 +136,13 @@ TEST(TensorTest, EmptyTensorTest) {
   auto& location = t.Location();
   ASSERT_STREQ(location.name, CPU);
   EXPECT_EQ(location.id, 0);
-  EXPECT_EQ(location.type, OrtAllocatorType::OrtArenaAllocator);
+
+  // arena is disabled for CPUExecutionProvider on x86 and JEMalloc
+#if (defined(__amd64__) || defined(_M_AMD64)) && !defined(USE_JEMALLOC)
+  EXPECT_EQ(location.alloc_type, OrtAllocatorType::OrtArenaAllocator);
+#else
+  EXPECT_EQ(location.alloc_type, OrtAllocatorType::OrtDeviceAllocator);
+#endif
 }
 
 TEST(TensorTest, StringTensorTest) {

From aa37dea5986280d1f12f4e3ecc12d44820401b63 Mon Sep 17 00:00:00 2001
From: Dmitri Smirnov <yuslepukhin@users.noreply.github.com>
Date: Tue, 14 Jan 2020 10:32:06 -0800
Subject: [PATCH 57/75] Convert ExternalProject Featurizers into git submodule
 (#2834)

Add git submodule for Featurizer library.
  Update cmake to build for git submodule.
---
 .gitmodules                       |  3 +++
 cmake/external/FeaturizersLibrary |  1 +
 cmake/external/featurizers.cmake  | 40 ++++---------------------------
 3 files changed, 9 insertions(+), 35 deletions(-)
 create mode 160000 cmake/external/FeaturizersLibrary

diff --git a/.gitmodules b/.gitmodules
index 19ec40f0ed5b5..bbe9507627428 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -46,3 +46,6 @@
 [submodule "server/external/spdlog"]
 	path = server/external/spdlog
 	url = https://github.com/gabime/spdlog.git
+[submodule "cmake/external/FeaturizersLibrary"]
+	path = cmake/external/FeaturizersLibrary
+	url = https://github.com/microsoft/FeaturizersLibrary.git
diff --git a/cmake/external/FeaturizersLibrary b/cmake/external/FeaturizersLibrary
new file mode 160000
index 0000000000000..ebec32ef06859
--- /dev/null
+++ b/cmake/external/FeaturizersLibrary
@@ -0,0 +1 @@
+Subproject commit ebec32ef06859b6399bf8854f18b91158c87760b
diff --git a/cmake/external/featurizers.cmake b/cmake/external/featurizers.cmake
index 5f1c49b73b2f7..5c10ee15a54c0 100644
--- a/cmake/external/featurizers.cmake
+++ b/cmake/external/featurizers.cmake
@@ -2,48 +2,17 @@
 # Licensed under the MIT License.
 # This source code should not depend on the onnxruntime and may be built independently
 
-set(featurizers_URL "https://github.com/microsoft/FeaturizersLibrary.git")
-set(featurizers_TAG "ebec32ef06859b6399bf8854f18b91158c87760b")
-
 set(featurizers_pref FeaturizersLibrary)
 set(featurizers_ROOT ${PROJECT_SOURCE_DIR}/external/${featurizers_pref})
 set(featurizers_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR}/external/${featurizers_pref})
 
-# Windows required git config setting
-# and an a switch whether we link to MSVCRT statically or dynamically
-if (WIN32)
-    ExternalProject_Add(featurizers_lib
-            PREFIX ${featurizers_pref}
-            GIT_REPOSITORY ${featurizers_URL}
-            GIT_TAG ${featurizers_TAG}
-            # Need this to properly checkout crlf
-            GIT_CONFIG core.autocrlf=input
-            SOURCE_DIR ${featurizers_ROOT}
-            # Location of CMakeLists.txt
-            SOURCE_SUBDIR src/Featurizers
-            BINARY_DIR ${featurizers_BINARY_DIR}
-            CMAKE_ARGS -Dfeaturizers_MSVC_STATIC_RUNTIME=${onnxruntime_MSVC_STATIC_RUNTIME}
-            INSTALL_COMMAND ""
-            
-        )
-else()
-    ExternalProject_Add(featurizers_lib
-            PREFIX ${featurizers_pref}
-            GIT_REPOSITORY ${featurizers_URL}
-            GIT_TAG ${featurizers_TAG}
-            SOURCE_DIR ${featurizers_ROOT}
-            # Location of CMakeLists.txt
-            SOURCE_SUBDIR src/Featurizers
-            BINARY_DIR ${featurizers_BINARY_DIR}
-            CMAKE_ARGS -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-            INSTALL_COMMAND ""
-        )
-endif()
+add_subdirectory(external/FeaturizersLibrary/src/Featurizers ${featurizers_BINARY_DIR} EXCLUDE_FROM_ALL)
+set_target_properties(FeaturizersCode PROPERTIES FOLDER "External/FeaturizersLibrary")
 
 add_library(onnxruntime_featurizers STATIC IMPORTED)
-add_dependencies(onnxruntime_featurizers featurizers_lib)
-target_include_directories(onnxruntime_featurizers INTERFACE ${featurizers_ROOT}/src)
+add_dependencies(onnxruntime_featurizers FeaturizersCode)
 
+target_include_directories(onnxruntime_featurizers INTERFACE ${featurizers_ROOT}/src)
 if(MSVC)
   set_property(TARGET onnxruntime_featurizers PROPERTY IMPORTED_LOCATION
     ${CMAKE_CURRENT_BINARY_DIR}/external/${featurizers_pref}/${CMAKE_BUILD_TYPE}/FeaturizersCode.lib)
@@ -52,6 +21,7 @@ else()
     ${CMAKE_CURRENT_BINARY_DIR}/external/${featurizers_pref}/libFeaturizersCode.a)
 endif()
 
+
 if (WIN32)
     # Add Code Analysis properties to enable C++ Core checks. Have to do it via a props file include.
     set_target_properties(onnxruntime_featurizers PROPERTIES VS_USER_PROPS ${PROJECT_SOURCE_DIR}/ConfigureVisualStudioCodeAnalysis.props)

From 8643f3ebbb7d52952d4a356ea25b2e5375cedf51 Mon Sep 17 00:00:00 2001
From: Ashwini Khade <askhade@microsoft.com>
Date: Tue, 14 Jan 2020 11:15:50 -0800
Subject: [PATCH 58/75] add domain check for nodes + update documentation
 (#2831)

---
 onnxruntime/python/tools/quantization/README.md | 17 ++++++++++++++++-
 .../python/tools/quantization/quantize.py       |  8 ++++++--
 2 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/onnxruntime/python/tools/quantization/README.md b/onnxruntime/python/tools/quantization/README.md
index 37d992c0b2a3e..c2da60a9aba8b 100644
--- a/onnxruntime/python/tools/quantization/README.md
+++ b/onnxruntime/python/tools/quantization/README.md
@@ -1,6 +1,6 @@
 # Quantization tool Overview
 This tool supports 8 bit linear quantization of an onnx model. quantize() takes a model in ModelProto format and returns the quantized model in ModelProto format.
-Today ORT does not guarantee support for E2E model quantization, meaning all ONNX ops do not have support for 8 bit data types therefore only the supported ops in the model are quantized. For rest of the ops inputs are reconverted to FP32.
+Today ORT does not guarantee support for E2E model quantization, meaning since not all ONNX ops have support for 8 bit data types therefore only the supported ops in the model are quantized. For rest of the ops inputs are reconverted to FP32.
 
 List of Supported Quantized Ops:
 The following ops were chosen as phase 1 ops because in most of the CNN models these ops consume most amount of compute and power and therefore there is benefit in quantizing these ops to get perf benefits.
@@ -25,6 +25,21 @@ The following ops were chosen as phase 1 ops because in most of the CNN models t
 
  Zero point represents zero in quantization space. It is important that floating point zero value be exactly representable in quantization space. This is because in lot of CNNs, zero padding is used and if after quantization it is not possible to represent 0 uniquely then it will lead to accuracy errors.
 
+## Quantization and model opset versions
+Quantization is fairly new in ONNX and ONNXRuntime. Quantization ops were introduced in ONNX opset version 10. Therefore it is important that the model which is being quantized be opset 10 or higher. In case the model opset version is < 10 then it is recommended that the model should be reconverted to ONNX from its original framework using the latest opset.
+
+Quantization tool displays a warning when the model opset version is < 10 and still goes ahead and quantizes the model and at the end changes the opset version to 10. It is the responsibility of the model owner to run model checker and make sure the model is valid. If the model is not valid then use the above recommended way i.e. reconvert the model from original framework.
+
+## Quantization and Graph Optimization
+Please note quantization and graph optimizations may not always work together.
+
+### Quantizing an optimized model
+If a model is optimized using level 99 (i.e. all possible optimizations are run on that model) then it is possible that after these optimizations are applied the model is converted in a way that quantization cannot be applied on this model anymore and therefore after running quantization script there will be no change in the model. 
+
+### Optimizing a quantized model
+Same goes other way round. After quantizing a model some graph optimizations which otherwise might have been applicable on this model may not be applicable anymore. 
+
+It is advised that the model owner be aware of this and run perf evaluations to understand which technique gives the best performance for their model.
 
 ## Quantize an ONNX model
 ```python
diff --git a/onnxruntime/python/tools/quantization/quantize.py b/onnxruntime/python/tools/quantization/quantize.py
index e886fd0df9923..1d5b393c10205 100644
--- a/onnxruntime/python/tools/quantization/quantize.py
+++ b/onnxruntime/python/tools/quantization/quantize.py
@@ -262,8 +262,12 @@ def quantize_model(self):
         # Create a new topologically sorted list for quantizing a model
         new_list = []
         for node in self.model.graph.node:
+            # if a list of ops to be quantized is provided then only quantize those ops
             if self.nodes_to_quantize is not None and node.name not in self.nodes_to_quantize:
                 new_list +=self._handle_other_ops(node, new_list)
+            # only onnx domain ops can be quantized today
+            elif node.domain != "ai.onnx" or node.domain != "":
+                new_list +=self._handle_other_ops(node, new_list)
             else:
                 if node.op_type == 'Conv':
                     new_list += self._quantize_convolution(node, new_list)
@@ -274,7 +278,7 @@ def quantize_model(self):
                 elif node.op_type == 'Relu' or node.op_type == 'Clip':
                     new_list +=self._handle_activation_ops(node, new_list)
                 else:
-                    new_list +=self._handle_other_ops(node, new_list)                    
+                    new_list +=self._handle_other_ops(node, new_list)
 
         # extend is used to append to the list for a protobuf fields
         # https://developers.google.com/protocol-buffers/docs/reference/python-generated?csw=1#fields
@@ -284,7 +288,7 @@ def quantize_model(self):
         # Remove weights which are already quantized from graph.
         self._remove_quantized_weights()
 
-        # update opset.
+        # update opset. 
         opset_info = next((opset for opset in self.model.opset_import if opset.domain == '' or opset.domain == onnx_domain), None)
         if opset_info is not None:
             self.model.opset_import.remove(opset_info)

From db05436fc02a8b07b9bfc5e72e9ea5273a484f5b Mon Sep 17 00:00:00 2001
From: Ori Levari <ori.levari@microsoft.com>
Date: Wed, 8 Jan 2020 15:04:59 -0800
Subject: [PATCH 59/75] User/orilevari/32bit comparison warning (#2800)

* use correct type for for loop

* explicitly specify void for parameters of OrtGetApiBase because the function is defined in c, so when the function is just (), it is interpreted as having an unknown number of parameters. This was causing compiler warning C4276.
---
 include/onnxruntime/core/session/onnxruntime_c_api.h | 2 +-
 onnxruntime/core/session/onnxruntime_c_api.cc        | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/include/onnxruntime/core/session/onnxruntime_c_api.h b/include/onnxruntime/core/session/onnxruntime_c_api.h
index 247d204f9b0de..176b988ad0f3b 100644
--- a/include/onnxruntime/core/session/onnxruntime_c_api.h
+++ b/include/onnxruntime/core/session/onnxruntime_c_api.h
@@ -218,7 +218,7 @@ struct OrtApiBase {
 };
 typedef struct OrtApiBase OrtApiBase;
 
-ORT_EXPORT const OrtApiBase* ORT_API_CALL OrtGetApiBase() NO_EXCEPTION;
+ORT_EXPORT const OrtApiBase* ORT_API_CALL OrtGetApiBase(void) NO_EXCEPTION;
 
 struct OrtApi {
   /**
diff --git a/onnxruntime/core/session/onnxruntime_c_api.cc b/onnxruntime/core/session/onnxruntime_c_api.cc
index 99ff0212ad8ab..de18d1fe64fdf 100644
--- a/onnxruntime/core/session/onnxruntime_c_api.cc
+++ b/onnxruntime/core/session/onnxruntime_c_api.cc
@@ -1460,7 +1460,7 @@ ORT_API(const char*, OrtApis::GetVersionString) {
   return ORT_VERSION;
 }
 
-const OrtApiBase* ORT_API_CALL OrtGetApiBase() NO_EXCEPTION {
+const OrtApiBase* ORT_API_CALL OrtGetApiBase(void) NO_EXCEPTION {
   return &ort_api_base;
 }
 

From cff266e1b9f10b8af6cefa72a572fd219c3398cb Mon Sep 17 00:00:00 2001
From: Tiago Koji Castro Shibata <tiago.shibata@gmail.com>
Date: Thu, 2 Jan 2020 15:34:48 -0800
Subject: [PATCH 60/75] Fix cgmanifest.json generating script (#2770)

* Fix protobuf submodule name

* Workaround pygit2 bug
---
 .gitmodules                    |  2 +-
 tools/python/get_submodules.py | 41 +++++++++++++++++++++++++++++-----
 2 files changed, 37 insertions(+), 6 deletions(-)

diff --git a/.gitmodules b/.gitmodules
index bbe9507627428..ee87ea533ebac 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,4 +1,4 @@
-[submodule "external/protobuf"]
+[submodule "cmake/external/protobuf"]
 	path = cmake/external/protobuf
 	url = https://github.com/google/protobuf.git
 [submodule "cmake/external/googletest"]
diff --git a/tools/python/get_submodules.py b/tools/python/get_submodules.py
index 6c86048367291..7f4ce58b3a5f0 100644
--- a/tools/python/get_submodules.py
+++ b/tools/python/get_submodules.py
@@ -1,12 +1,44 @@
-import pygit2
+from operator import attrgetter
+from pathlib import Path
 import argparse
+import configparser
 import json
+import re
+
+import pygit2
+
 
 def format_component(submod):
-    return {"component":{"type":"git","git":{"commitHash":str(submod.head_id), "repositoryUrl":str(submod.url)}}}
+    return {"component": {"type": "git", "git": {"commitHash": str(submod.head_id), "repositoryUrl": submod.url}}}
+
+def lookup_submodule(repo, submodule_path):
+    submodule = repo.lookup_submodule(submodule_path)
+    try:
+        # Some submodules have names which don't correspond to the actual path in the repo
+        # (e.g. 'git submodule init' was called with the --name option, or the submodule
+        # was moved and the old name was kept). listall_submodules() returns submodule paths,
+        # but pygit up to 1.0.1 requires the submodule name (not the path) in lookup_submodule
+        # to be able to access the URL and other properties.
+        # This seems to be a bug in pygit2, since its documentation says the submodules can
+        # be opened by path.
+        # If accessing the URL throws a RuntimeError, we get the submodule name manually from
+        # .gitmodules.
+        submodule.url
+        return submodule
+    except RuntimeError:
+        pass
+
+    config = configparser.ConfigParser()
+    config.read(Path(repo.workdir, '.gitmodules'))
+    for section in config.sections():
+        if config[section]['path'] == submodule_path:
+            name = re.fullmatch('submodule "(.*)"', section).group(1)
+            submodule = repo.lookup_submodule(name)
+            return submodule
+    raise NotImplementedError()  # should not be reached
 
 def process_component(repo):
-    return [repo.lookup_submodule(submod) for submod in repo.listall_submodules()]
+    return [lookup_submodule(repo, submod) for submod in repo.listall_submodules()]
 
 def recursive_process(base_repo):
     processed_subs = []
@@ -24,10 +56,9 @@ def main(repo_path, output_file):
     with open(output_file, 'w') as f:
         json.dump(registrations, f, indent=4, sort_keys=True)
 
-if __name__=="__main__":
+if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     parser.add_argument("base_repository", help="path to base repository to get registrations for.")
     parser.add_argument("-o", "--output", help="output file name.", default="cgmanifest.json")
     args = parser.parse_args()
     main(args.base_repository, args.output)
-

From dbe7d97fa1ab155f1309bced87199527e8f35bd2 Mon Sep 17 00:00:00 2001
From: Tiago Koji Castro Shibata <tiago.shibata@gmail.com>
Date: Tue, 14 Jan 2020 15:03:33 -0800
Subject: [PATCH 61/75] CMake cross-generator fixes (#2790)

* Fix compilation w/ non-VS CMake generators

* Fix custom WINMD target in Ninja

* Remove usage of msbuild .targets file

* Fix linking using DML in Ninja

* Automate SDK kit version choice

* Cleanup DML package install

* Fix SDK version detection

* Fix comment

* Revert unittest linkage changes

* Fix latest SDK detection

* Don't link to non-uapcore libraries

* Remove MessageBoxA reference and unused link libs
---
 cmake/external/dml.cmake                      |  8 +--
 cmake/onnxruntime_common.cmake                | 16 +++++
 cmake/onnxruntime_mlas.cmake                  |  6 +-
 cmake/onnxruntime_providers.cmake             | 35 +++++++---
 cmake/onnxruntime_unittests.cmake             | 12 ++--
 cmake/precompiled_header.cmake                | 15 ++--
 cmake/winml.cmake                             | 25 ++++---
 cmake/winml_cppwinrt.cmake                    | 69 ++++++++++---------
 cmake/winml_sdk_helpers.cmake                 | 24 +++++--
 .../core/platform/windows/debug_alloc.cc      |  8 +--
 10 files changed, 135 insertions(+), 83 deletions(-)

diff --git a/cmake/external/dml.cmake b/cmake/external/dml.cmake
index 41de28fdd2e0a..35c606f6412e8 100644
--- a/cmake/external/dml.cmake
+++ b/cmake/external/dml.cmake
@@ -20,19 +20,17 @@ if (NOT onnxruntime_USE_CUSTOM_DIRECTML)
   set(NUGET_CONFIG ${PROJECT_SOURCE_DIR}/../NuGet.config)
   set(PACKAGES_CONFIG ${PROJECT_SOURCE_DIR}/../packages.config)
   set(PACKAGES_DIR ${CMAKE_CURRENT_BINARY_DIR}/packages)
+  set(DML_PACKAGE_DIR ${PACKAGES_DIR}/DirectML.0.0.1)
 
   # Restore nuget packages, which will pull down the DirectML redist package
   add_custom_command(
-    OUTPUT restore_packages.stamp
+    OUTPUT ${DML_PACKAGE_DIR}/bin/x64/DirectML.lib ${DML_PACKAGE_DIR}/bin/x86/DirectML.lib
     DEPENDS ${PACKAGES_CONFIG} ${NUGET_CONFIG}
     COMMAND ${CMAKE_CURRENT_BINARY_DIR}/nuget/src/nuget restore ${PACKAGES_CONFIG} -PackagesDirectory ${PACKAGES_DIR} -ConfigFile ${NUGET_CONFIG}
-    COMMAND ${CMAKE_COMMAND} -E touch restore_packages.stamp
     VERBATIM)
 
-  add_custom_target(RESTORE_PACKAGES ALL DEPENDS restore_packages.stamp)
+  add_custom_target(RESTORE_PACKAGES ALL DEPENDS ${DML_PACKAGE_DIR}/bin/x64/DirectML.lib ${DML_PACKAGE_DIR}/bin/x86/DirectML.lib)
   add_dependencies(RESTORE_PACKAGES nuget)
-
-  list(APPEND onnxruntime_EXTERNAL_DEPENDENCIES RESTORE_PACKAGES)
 else()
   include_directories(${dml_INCLUDE_DIR})
 endif()
diff --git a/cmake/onnxruntime_common.cmake b/cmake/onnxruntime_common.cmake
index 63a8283429dce..52aa2f3989d9d 100644
--- a/cmake/onnxruntime_common.cmake
+++ b/cmake/onnxruntime_common.cmake
@@ -44,6 +44,22 @@ else()
     endif()
 endif()
 
+if(CMAKE_GENERATOR_PLATFORM)
+    # Multi-platform generator
+    set(onnxruntime_target_platform ${CMAKE_GENERATOR_PLATFORM})
+else()
+    set(onnxruntime_target_platform ${CMAKE_SYSTEM_PROCESSOR})
+endif()
+if(onnxruntime_target_platform STREQUAL "ARM64")
+    set(onnxruntime_target_platform "ARM64")
+elseif(onnxruntime_target_platform STREQUAL "ARM" OR CMAKE_GENERATOR MATCHES "ARM")
+    set(onnxruntime_target_platform "ARM")
+elseif(onnxruntime_target_platform STREQUAL "x64" OR onnxruntime_target_platform STREQUAL "x86_64" OR onnxruntime_target_platform STREQUAL "AMD64" OR CMAKE_GENERATOR MATCHES "Win64")
+    set(onnxruntime_target_platform "x64")
+elseif(onnxruntime_target_platform STREQUAL "x86" OR onnxruntime_target_platform STREQUAL "i386" OR onnxruntime_target_platform STREQUAL "i686")
+    set(onnxruntime_target_platform "x86")
+endif()
+
 file(GLOB onnxruntime_common_src CONFIGURE_DEPENDS
     ${onnxruntime_common_src_patterns}
     )
diff --git a/cmake/onnxruntime_mlas.cmake b/cmake/onnxruntime_mlas.cmake
index e7b4213bc6cbb..38ec173dfc0b1 100644
--- a/cmake/onnxruntime_mlas.cmake
+++ b/cmake/onnxruntime_mlas.cmake
@@ -19,7 +19,7 @@ set(mlas_common_srcs
 )
 
 if(MSVC)
-  if(CMAKE_GENERATOR_PLATFORM STREQUAL "ARM64")
+  if(onnxruntime_target_platform STREQUAL "ARM64")
     set(asm_filename ${ONNXRUNTIME_ROOT}/core/mlas/lib/arm64/SgemmKernelNeon.asm)
     set(pre_filename ${CMAKE_CURRENT_BINARY_DIR}/SgemmKernelNeon.i)
     set(obj_filename ${CMAKE_CURRENT_BINARY_DIR}/SgemmKernelNeon.obj)
@@ -38,11 +38,11 @@ if(MSVC)
             armasm64.exe ${ARMASM_FLAGS} ${pre_filename} ${obj_filename}
     )
     set(mlas_platform_srcs ${obj_filename})
-  elseif(CMAKE_GENERATOR_PLATFORM STREQUAL "ARM" OR CMAKE_GENERATOR MATCHES "ARM")
+  elseif(onnxruntime_target_platform STREQUAL "ARM")
     set(mlas_platform_srcs
       ${ONNXRUNTIME_ROOT}/core/mlas/lib/arm/sgemmc.cpp
     )
-  elseif(CMAKE_GENERATOR_PLATFORM STREQUAL "x64" OR CMAKE_GENERATOR MATCHES "Win64")
+  elseif(onnxruntime_target_platform STREQUAL "x64")
     enable_language(ASM_MASM)
 
     set(mlas_platform_srcs
diff --git a/cmake/onnxruntime_providers.cmake b/cmake/onnxruntime_providers.cmake
index 9875955e3cd89..1b7047318ee85 100644
--- a/cmake/onnxruntime_providers.cmake
+++ b/cmake/onnxruntime_providers.cmake
@@ -217,7 +217,7 @@ if (onnxruntime_USE_TENSORRT)
   if ( CMAKE_COMPILER_IS_GNUCC )
     set(CMAKE_CXX_FLAGS  "${CMAKE_CXX_FLAGS} -Wno-unused-parameter -Wno-missing-field-initializers")
   endif()
-  set(CXX_VERSION_DEFINED TRUE)  
+  set(CXX_VERSION_DEFINED TRUE)
   add_subdirectory(${ONNXRUNTIME_ROOT}/../cmake/external/onnx-tensorrt)
   set(CMAKE_CXX_FLAGS ${OLD_CMAKE_CXX_FLAGS})
   if (WIN32)
@@ -303,7 +303,7 @@ if (onnxruntime_USE_OPENVINO)
     if(WIN32)
      set(OPENVINO_LIB_DIR $ENV{INTEL_OPENVINO_DIR}/deployment_tools/inference_engine/lib/intel64/Release)
      set(OPENVINO_TBB_DIR $ENV{INTEL_OPENVINO_DIR}/deployment_tools/inference_engine/lib/intel64/Release)
-     set(OPENVINO_MKL_TINY_DIR $ENV{INTEL_OPENVINO_DIR}/deployment_tools/inference_engine/bin/intel64/Release)	 
+     set(OPENVINO_MKL_TINY_DIR $ENV{INTEL_OPENVINO_DIR}/deployment_tools/inference_engine/bin/intel64/Release)
     else()
      set(OPENVINO_LIB_DIR $ENV{INTEL_OPENVINO_DIR}/deployment_tools/inference_engine/lib/intel64/)
      set(OPENVINO_TBB_DIR $ENV{INTEL_OPENVINO_DIR}/deployment_tools/inference_engine/external/tbb/lib)
@@ -327,9 +327,9 @@ if (onnxruntime_USE_OPENVINO)
   else()
     target_include_directories(onnxruntime_providers_openvino SYSTEM PUBLIC ${ONNXRUNTIME_ROOT} ${eigen_INCLUDE_DIRS} ${OPENVINO_INCLUDE_DIR} ${OPENVINO_EXTENSIONS_DIR} ${OPENVINO_LIB_DIR} ${OPENVINO_TBB_INCLUDE_DIR} ${PYTHON_INCLUDE_DIRS})
   endif()
-  
-   if (WIN32)   
-     string(REPLACE "include" "libs" PYTHON_LIB ${PYTHON_INCLUDE_DIRS})	
+
+   if (WIN32)
+     string(REPLACE "include" "libs" PYTHON_LIB ${PYTHON_INCLUDE_DIRS})
 	   find_package(InferenceEngine 2.1 REQUIRED)
      set(PYTHON_LIBRARIES ${PYTHON_LIB})
      set(OPENVINO_CPU_EXTENSION_DIR ${onnxruntime_BINARY_DIR}/ie_cpu_extension/${CMAKE_BUILD_TYPE})
@@ -430,22 +430,37 @@ if (onnxruntime_USE_DML)
   onnxruntime_add_include_to_target(onnxruntime_providers_dml onnxruntime_common onnxruntime_framework onnx onnx_proto protobuf::libprotobuf)
   add_dependencies(onnxruntime_providers_dml ${onnxruntime_EXTERNAL_DEPENDENCIES})
   target_include_directories(onnxruntime_providers_dml PRIVATE ${ONNXRUNTIME_ROOT} ${ONNXRUNTIME_ROOT}/../cmake/external/wil/include)
-  
-  target_link_libraries(onnxruntime_providers_dml ${CMAKE_CURRENT_BINARY_DIR}/packages/DirectML.0.0.1/build/DirectML.targets)
-  target_link_libraries(onnxruntime_providers_dml d3d12.lib dxgi.lib)
+
+  if(NOT onnxruntime_target_platform STREQUAL "x86" AND NOT onnxruntime_target_platform STREQUAL "x64")
+    message(FATAL_ERROR "Target platform ${onnxruntime_target_platform} is not supported by DML")
+  endif()
+  foreach(file "DirectML.dll" "DirectML.pdb" "DirectML.Debug.dll" "DirectML.Debug.pdb")
+    add_custom_command(TARGET onnxruntime_providers_dml
+      POST_BUILD
+      COMMAND ${CMAKE_COMMAND} -E copy_if_different
+        "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/${file}" $<TARGET_FILE_DIR:onnxruntime_providers_dml>)
+  endforeach()
+
+  function(target_add_dml target)
+    target_link_libraries(${target} PRIVATE "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/DirectML.lib")
+    target_include_directories(${target} PRIVATE "${DML_PACKAGE_DIR}/include")
+  endfunction()
+
+  target_add_dml(onnxruntime_providers_dml)
+  target_link_libraries(onnxruntime_providers_dml PRIVATE d3d12.lib dxgi.lib delayimp.lib)
   list(APPEND ONNXRUNTIME_LINKER_FLAGS "/DELAYLOAD:DirectML.dll /DELAYLOAD:d3d12.dll /DELAYLOAD:dxgi.dll")
 
   # The DML EP requires C++17
   set_target_properties(onnxruntime_providers_dml PROPERTIES CXX_STANDARD 17)
   set_target_properties(onnxruntime_providers_dml PROPERTIES CXX_STANDARD_REQUIRED ON)
-  
+
   target_compile_definitions(onnxruntime_providers_dml PRIVATE ONNX_NAMESPACE=onnx ONNX_ML LOTUS_LOG_THRESHOLD=2 LOTUS_ENABLE_STDERR_LOGGING PLATFORM_WINDOWS)
   target_compile_definitions(onnxruntime_providers_dml PRIVATE UNICODE _UNICODE NOMINMAX)
   if (MSVC)
     target_compile_definitions(onnxruntime_providers_dml PRIVATE _SILENCE_CXX17_ITERATOR_BASE_CLASS_DEPRECATION_WARNING)
     target_compile_options(onnxruntime_providers_dml PRIVATE "/W3")
   endif()
-  
+
   install(DIRECTORY ${PROJECT_SOURCE_DIR}/../include/onnxruntime/core/providers/dml  DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/onnxruntime/core/providers)
 
   set_target_properties(onnxruntime_providers_dml PROPERTIES LINKER_LANGUAGE CXX)
diff --git a/cmake/onnxruntime_unittests.cmake b/cmake/onnxruntime_unittests.cmake
index 09aedbe005ae8..1cba4e472066a 100644
--- a/cmake/onnxruntime_unittests.cmake
+++ b/cmake/onnxruntime_unittests.cmake
@@ -44,7 +44,7 @@ function(AddTest)
   endif()
   if (onnxruntime_ENABLE_LANGUAGE_INTEROP_OPS AND onnxruntime_ENABLE_PYTHON)
     target_compile_definitions(${_UT_TARGET} PRIVATE ENABLE_LANGUAGE_INTEROP_OPS)
-  endif()  
+  endif()
   if (WIN32)
     if (onnxruntime_USE_CUDA)
       # disable a warning from the CUDA headers about unreferenced local functions
@@ -318,7 +318,7 @@ if (onnxruntime_USE_DNNL)
   target_compile_definitions(onnxruntime_test_utils_for_framework PUBLIC USE_DNNL=1)
 endif()
 if (onnxruntime_USE_DML)
-  target_link_libraries(onnxruntime_test_utils_for_framework PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/packages/DirectML.0.0.1/build/DirectML.targets)
+  target_add_dml(onnxruntime_test_utils_for_framework)
 endif()
 add_dependencies(onnxruntime_test_utils_for_framework ${onnxruntime_EXTERNAL_DEPENDENCIES})
 target_include_directories(onnxruntime_test_utils_for_framework PUBLIC "${TEST_SRC_DIR}/util/include" PRIVATE ${eigen_INCLUDE_DIRS} ${ONNXRUNTIME_ROOT})
@@ -336,7 +336,7 @@ if (onnxruntime_USE_DNNL)
   target_compile_definitions(onnxruntime_test_utils PUBLIC USE_DNNL=1)
 endif()
 if (onnxruntime_USE_DML)
-  target_link_libraries(onnxruntime_test_utils PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/packages/DirectML.0.0.1/build/DirectML.targets)
+  target_add_dml(onnxruntime_test_utils)
 endif()
 add_dependencies(onnxruntime_test_utils ${onnxruntime_EXTERNAL_DEPENDENCIES})
 target_include_directories(onnxruntime_test_utils PUBLIC "${TEST_SRC_DIR}/util/include" PRIVATE ${eigen_INCLUDE_DIRS} ${ONNXRUNTIME_ROOT})
@@ -722,7 +722,7 @@ if (onnxruntime_BUILD_SERVER)
       set_source_files_properties("${TEST_SRC_DIR}/server/unit_tests/executor_test.cc" PROPERTIES COMPILE_FLAGS -Wno-unused-parameter)
     endif()
   endif()
-  
+
   add_library(onnxruntime_test_utils_for_server ${onnxruntime_test_server_src})
   onnxruntime_add_include_to_target(onnxruntime_test_utils_for_server onnxruntime_test_utils_for_framework gtest gmock onnx onnx_proto server_proto server_grpc_proto)
   add_dependencies(onnxruntime_test_utils_for_server onnxruntime_server_lib onnxruntime_server_http_core_lib Boost ${onnxruntime_EXTERNAL_DEPENDENCIES})
@@ -751,13 +751,13 @@ if (onnxruntime_BUILD_SERVER)
           LANGUAGE python
           TARGET onnxruntime_server_tests
           OUT_VAR server_test_py)
-          
+
   set(grpc_py "${CMAKE_CURRENT_BINARY_DIR}/prediction_service_pb2_grpc.py")
 
   add_custom_command(
     TARGET onnxruntime_server_tests
     COMMAND $<TARGET_FILE:protobuf::protoc>
-    ARGS 
+    ARGS
       --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
       --plugin=protoc-gen-grpc="${_GRPC_PY_PLUGIN_EXECUTABLE}"
       -I ${grpc_proto_path}
diff --git a/cmake/precompiled_header.cmake b/cmake/precompiled_header.cmake
index f41310fac524c..3bebf7341809b 100644
--- a/cmake/precompiled_header.cmake
+++ b/cmake/precompiled_header.cmake
@@ -2,8 +2,7 @@
 # header name as input. The function will generate a .cpp file that includes the header and is used
 # to generate the precompiled header; this source file is added to the target's sources.
 function(target_precompiled_header target_name header_name)
-    if (MSVC)
-
+    if (MSVC AND CMAKE_VS_PLATFORM_TOOLSET)
         # The input precompiled header source (i.e. the '.h' file used for the precompiled header).
         set(pch_header_path ${header_name})
         get_filename_component(header_base_name ${header_name} NAME_WE)
@@ -14,14 +13,14 @@ function(target_precompiled_header target_name header_name)
         set(pch_source_content "// THIS FILE IS GENERATED BY CMAKE\n#include \"${pch_header_path}\"")
         file(WRITE ${pch_source_path} ${pch_source_content})
         set_source_files_properties(${pch_source_path} PROPERTIES COMPILE_FLAGS "/Yc${pch_header_path}")
-    
+
         # The target's C++ sources use the precompiled header (/Yu). Source-level properties will
-        # take precedence over target-level properties, so this will not change the generated source 
+        # take precedence over target-level properties, so this will not change the generated source
         # file's property to create the precompiled header (/Yc).
         target_compile_options(${target_name} PRIVATE $<$<COMPILE_LANGUAGE:CXX>:/Yu${header_name}>)
-    
+
         # Append generated precompiled source to target's sources.
         target_sources(${target_name} PRIVATE ${pch_source_path})
-    
-    endif(MSVC)
-endfunction()
\ No newline at end of file
+
+    endif()
+endfunction()
diff --git a/cmake/winml.cmake b/cmake/winml.cmake
index 71824b3997fcf..f5171158f230e 100644
--- a/cmake/winml.cmake
+++ b/cmake/winml.cmake
@@ -132,7 +132,7 @@ list(APPEND winml_adapter_files
     ${winml_adapter_dir}/WinMLAdapter.cpp
     ${winml_adapter_dir}/WinMLAdapter.h
     ${winml_adapter_dir}/ZeroCopyInputStreamWrapper.cpp
-    ${winml_adapter_dir}/ZeroCopyInputStreamWrapper.h    
+    ${winml_adapter_dir}/ZeroCopyInputStreamWrapper.h
     )
 
 if (onnxruntime_USE_DML)
@@ -159,6 +159,7 @@ add_dependencies(winml_adapter ${onnxruntime_EXTERNAL_DEPENDENCIES})
 target_precompiled_header(winml_adapter pch.h)
 
 # Includes
+target_include_directories(winml_adapter PRIVATE ${CMAKE_CURRENT_BINARY_DIR})                             # windows machine learning generated component headers
 target_include_directories(winml_adapter PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api)                   # windows machine learning generated component headers
 target_include_directories(winml_adapter PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api/comp_generated)    # windows machine learning generated component headers
 target_include_directories(winml_adapter PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml/sdk/cppwinrt/include)  # sdk cppwinrt headers
@@ -181,11 +182,11 @@ add_dependencies(winml_adapter winml_api_native_internal)
 # Link libraries
 target_link_libraries(winml_adapter PRIVATE wil)
 if (onnxruntime_USE_DML)
-  target_link_libraries(winml_adapter PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/packages/DirectML.0.0.1/build/DirectML.targets)
+  target_add_dml(winml_adapter)
 endif(onnxruntime_USE_DML)
 
 # add it to the onnxruntime shared library
-set(onnxruntime_winml windowsapp.lib winml_adapter)
+set(onnxruntime_winml winml_adapter)
 list(APPEND onnxruntime_EXTERNAL_DEPENDENCIES winml_adapter)
 
 ###########################
@@ -230,6 +231,7 @@ target_compile_definitions(winml_lib_image PRIVATE _SCL_SECURE_NO_WARNINGS)
 target_precompiled_header(winml_lib_image pch.h)
 
 # Includes
+target_include_directories(winml_lib_image PRIVATE ${CMAKE_CURRENT_BINARY_DIR})                                                               # windows machine learning generated component headers
 target_include_directories(winml_lib_image PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api)                                                     # windows machine learning generated component headers
 target_include_directories(winml_lib_image PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api/comp_generated)                                      # windows machine learning generated component headers
 target_include_directories(winml_lib_image PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml/sdk/cppwinrt/include)                                    # sdk cppwinrt headers
@@ -258,7 +260,7 @@ add_dependencies(winml_lib_image winml_api_native_internal)
 # Link libraries
 target_link_libraries(winml_lib_image PRIVATE wil)
 if (onnxruntime_USE_DML)
-  target_link_libraries(winml_lib_image PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/packages/DirectML.0.0.1/build/DirectML.targets)
+  target_add_dml(winml_lib_image)
 endif(onnxruntime_USE_DML)
 
 
@@ -360,7 +362,7 @@ add_dependencies(winml_lib_api winml_api_native_internal)
 # Link libraries
 target_link_libraries(winml_lib_api PRIVATE wil)
 if (onnxruntime_USE_DML)
-  target_link_libraries(winml_lib_api PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/packages/DirectML.0.0.1/build/DirectML.targets)
+  target_add_dml(winml_lib_api)
 endif(onnxruntime_USE_DML)
 
 
@@ -438,10 +440,17 @@ if (onnxruntime_USE_DML)
   set(delayload_dml "/DELAYLOAD:directml.dll")
 endif(onnxruntime_USE_DML)
 
+# The default libraries to link with in Windows are kernel32.lib;user32.lib;gdi32.lib;winspool.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;comdlg32.lib;advapi32.lib
+# Remove them and use the onecore umbrella library instead
+set(CMAKE_C_STANDARD_LIBRARIES "onecoreuap_apiset.lib")
+set(CMAKE_CXX_STANDARD_LIBRARIES "onecoreuap_apiset.lib")
+foreach(default_lib kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdgl32.lib advapi32.lib)
+  set(removed_libs "${removed_libs} /NODEFAULTLIB:${default_lib}")
+endforeach()
 set_target_properties(winml_dll
     PROPERTIES
     LINK_FLAGS
-    "/DEF:${WINML_DIR}/windows.ai.machinelearning.def ${os_component_link_flags} /DELAYLOAD:d3d12.dll /DELAYLOAD:d3d11.dll /DELAYLOAD:dxgi.dll ${delayload_dml}")
+    "/DEF:${WINML_DIR}/windows.ai.machinelearning.def ${os_component_link_flags} /DELAYLOAD:d3d12.dll /DELAYLOAD:d3d11.dll /DELAYLOAD:dxgi.dll ${delayload_dml} ${removed_libs}")
 
 
 set_target_properties(winml_dll
@@ -467,11 +476,10 @@ endif("${CMAKE_BUILD_TYPE}" STREQUAL "Debug")
 target_link_libraries(winml_dll PRIVATE onnxruntime)
 target_link_libraries(winml_dll PRIVATE re2)
 target_link_libraries(winml_dll PRIVATE wil)
-#target_link_libraries(winml_dll PRIVATE windowsapp.lib)
 target_link_libraries(winml_dll PRIVATE winml_lib_api)
 target_link_libraries(winml_dll PRIVATE winml_lib_image)
 target_link_libraries(winml_dll PRIVATE winml_lib_telemetry)
-target_link_libraries(winml_dll PRIVATE onecoreuap_apiset.lib)
+target_link_libraries(winml_dll PRIVATE delayimp.lib)
 target_link_libraries(winml_dll PRIVATE ${DBGHELP})
 
 # 1 of 3 projects that fail in link with 'failed to do memory mapped file I/O' (Only release)
@@ -483,6 +491,7 @@ if("${CMAKE_BUILD_TYPE}" STREQUAL "Release")
   set_target_properties(winml_dll PROPERTIES VS_GLOBAL_PreferredToolArchitecture "x64")
 endif("${CMAKE_BUILD_TYPE}" STREQUAL "Release")
 
+option(onnxruntime_BUILD_WINML_TESTS "Build WinML tests" ON)
 if (onnxruntime_BUILD_WINML_TESTS)
   include(winml_unittests.cmake)
 endif()
diff --git a/cmake/winml_cppwinrt.cmake b/cmake/winml_cppwinrt.cmake
index 2d53ade270b2b..ea96a1b289764 100644
--- a/cmake/winml_cppwinrt.cmake
+++ b/cmake/winml_cppwinrt.cmake
@@ -102,7 +102,7 @@ function(target_cppwinrt
         # Get directory
         get_filename_component(idl_source_directory ${file} DIRECTORY)
 
-	set(target_outputs ${CMAKE_CURRENT_BINARY_DIR}/${target_name})
+        set(target_outputs ${CMAKE_CURRENT_BINARY_DIR}/${target_name})
         convert_forward_slashes_to_back(${target_outputs}/comp output_dir_back_slash)
         convert_forward_slashes_to_back(${target_outputs}/temp temp_dir_back_slash)
         convert_forward_slashes_to_back(${target_outputs}/comp_generated generated_dir_back_slash)
@@ -126,50 +126,53 @@ function(target_cppwinrt
                 /tlb ${tlb_filename}
                 ${idl_file_forward_slash}
             COMMAND
-                    ${cppwinrt_exe} -in \"${winmd_filename}\" -comp \"${output_dir_back_slash}\" -ref \"${sdk_metadata_directory}\" -out \"${generated_dir_back_slash}\" -verbose
+                    ${cppwinrt_exe} -in ${winmd_filename} -comp ${output_dir_back_slash} -ref ${sdk_metadata_directory} -out ${generated_dir_back_slash} -verbose
             COMMAND
                     # copy the generated component files into a temporary directory where headers exclusions will be applied
-                    xcopy \"${output_dir_back_slash}\" \"${temp_dir_back_slash}\\\" /Y /D
+                    xcopy ${output_dir_back_slash} ${temp_dir_back_slash}\\ /Y /D
             COMMAND
                     # for each file in the temp directory, ensure it is not in the exclusions list.
                     # if it is, then we need to delete it.
-                    for /f %%I in ('dir /b \"${temp_dir_back_slash}\"')
-                    do
-                    (
-                        for /f %%E in (${CPPWINRT_COMPONENT_EXCLUSION_LIST})
-                        do
-                        (
-                            if %%E == %%I
-                            (
-                                del \"${temp_dir_back_slash}\\%%I\"
-                            )
-                        )
-                    )
+                    cmd /C "@echo off \
+                    for /f %I in ('dir /b ${temp_dir_back_slash}') \
+                    do \
+                    ( \
+                        for /f %E in (${CPPWINRT_COMPONENT_EXCLUSION_LIST}) \
+                        do \
+                        ( \
+                            if %E == %I \
+                            ( \
+                                del ${temp_dir_back_slash}\\%I \
+                            ) \
+                        ) \
+                    )"
             COMMAND
                     # for each file in the temp directory, copy the file back into the source tree
                     # unless the file already exists
-                    for /f %%I in ('dir /b \"${temp_dir_back_slash}\"')
-                    do
-                    (
-                        if not exist \"${out_sources_folder}\\%%I\"
-                        (
-                            xcopy \"${temp_dir_back_slash}\\%%I\" \"${out_sources_folder}\\%%I\"
-                        )
-                    )
+                    cmd /C "@echo off \
+                    for /f %I in ('dir /b ${temp_dir_back_slash}') \
+                    do \
+                    ( \
+                        if not exist ${out_sources_folder}\\%I \
+                        ( \
+                            copy ${temp_dir_back_slash}\\%I ${out_sources_folder}\\%I \
+                        ) \
+                    )"
             COMMAND
                     # open the generated module.g.cpp and strip all the includes (lines) containing excluded headers
                     # write the new file out to module.g.excl.cpp.
-                    powershell -Command \"& {
-                        $exclusions = get-content '${CPPWINRT_COMPONENT_EXCLUSION_LIST}'\;
-                        (get-content '${module_g_cpp_back_slash}')
-                        | where {
-                            $str = $_\;
-                            $matches = ($exclusions | where { $str -match $_ }) \;
-                            $matches.Length -eq 0 }
-                        | Out-File '${module_g_ecxl_cpp_back_slash}'
-                    }\"
+                    powershell -Command "& { \
+                        $exclusions = get-content '${CPPWINRT_COMPONENT_EXCLUSION_LIST}'; \
+                        (get-content '${module_g_cpp_back_slash}') \
+                        | where { \
+                            $str = $_; \
+                            $matches = ($exclusions | where { $str -match $_ }); \
+                            $matches.Length -eq 0 } \
+                        | Out-File '${module_g_ecxl_cpp_back_slash}' \
+                    }"
             BYPRODUCTS
                     ${generated_dir_back_slash}/module.g.excl.cpp
+            VERBATIM
         )
 
         add_custom_target(
@@ -214,4 +217,4 @@ function(add_generate_cppwinrt_sdk_headers_target
 
         set_target_properties(${target_name} PROPERTIES FOLDER ${folder_name})
     endif()
-endfunction()
\ No newline at end of file
+endfunction()
diff --git a/cmake/winml_sdk_helpers.cmake b/cmake/winml_sdk_helpers.cmake
index 03a87653fe986..df9db93a6b467 100644
--- a/cmake/winml_sdk_helpers.cmake
+++ b/cmake/winml_sdk_helpers.cmake
@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.0)
 
-# utility 
+# utility
 function(convert_forward_slashes_to_back input output)
     string(REGEX REPLACE "/" "\\\\" backwards ${input})
     set(${output} ${backwards} PARENT_SCOPE)
@@ -16,7 +16,23 @@ function(get_installed_sdk
     set(${sdk_folder} ${win10_sdk_root} PARENT_SCOPE)
 
     # return the sdk version
-    set(${output_sdk_version} ${CMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION} PARENT_SCOPE)
+    if(CMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION)
+        set(${output_sdk_version} ${CMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION} PARENT_SCOPE)
+    else()
+        # choose the SDK matching the system version, or fallback to the latest
+        file(GLOB win10_sdks RELATIVE "${win10_sdk_root}/UnionMetadata" "${win10_sdk_root}/UnionMetadata/*.*.*.*")
+        list(GET win10_sdks 0 latest_sdk)
+        foreach(sdk IN LISTS win10_sdks)
+            string(FIND ${sdk} ${CMAKE_SYSTEM_VERSION} is_system_version)
+            if(NOT ${is_system_version} EQUAL -1)
+                set(${output_sdk_version} ${sdk} PARENT_SCOPE)
+                return()
+            elseif(sdk VERSION_GREATER latest_sdk)
+                set(latest_sdk ${sdk})
+            endif()
+        endforeach()
+        set(${output_sdk_version} ${latest_sdk} PARENT_SCOPE)
+    endif()
 endfunction()
 
 # current sdk binary directory
@@ -95,7 +111,7 @@ function(get_sdk
     set(${output_sdk_version} ${winml_WINDOWS_SDK_VERSION_OVERRIDE} PARENT_SCOPE)
   else()
     message(
-      FATAL_ERROR 
+      FATAL_ERROR
       "Options winml_WINDOWS_SDK_DIR_OVERRIDE and winml_WINDOWS_SDK_VERSION_OVERRIDE must be defined together, or not at all.")
   endif()
-endfunction()
\ No newline at end of file
+endfunction()
diff --git a/onnxruntime/core/platform/windows/debug_alloc.cc b/onnxruntime/core/platform/windows/debug_alloc.cc
index 83c38baa05248..57267f146d884 100644
--- a/onnxruntime/core/platform/windows/debug_alloc.cc
+++ b/onnxruntime/core/platform/windows/debug_alloc.cc
@@ -232,12 +232,8 @@ Memory_LeakCheck::~Memory_LeakCheck() {
     _snprintf_s(buffer, _TRUNCATE, "%d bytes of memory leaked in %d allocations", leaked_bytes, leak_count);
     string.append(buffer);
 
-    // If we're being actively debugged, show a message box to get the dev's attention
-    if (IsDebuggerPresent())
-      MessageBoxA(nullptr, string.c_str(), "Warning", MB_OK | MB_ICONWARNING);
-    else {
-      // If we're on the command line (like on a build machine), output to the console and exit(-1)
-      std::cout << "\n----- MEMORY LEAKS: " << string.c_str() << "\n";
+    std::cout << "\n----- MEMORY LEAKS: " << string.c_str() << "\n";
+    if (!IsDebuggerPresent()) {
       exit(-1);
     }
 

From 56030f8d74c55c05b3a5db918951939f9b32678b Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Tue, 14 Jan 2020 16:23:05 -0800
Subject: [PATCH 62/75] Fix Linux CUDA nuget packaging pipeline break

---
 onnxruntime/test/framework/cuda/fence_cuda_test.cc | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/onnxruntime/test/framework/cuda/fence_cuda_test.cc b/onnxruntime/test/framework/cuda/fence_cuda_test.cc
index b2f55064deaa5..c2c057a96402c 100644
--- a/onnxruntime/test/framework/cuda/fence_cuda_test.cc
+++ b/onnxruntime/test/framework/cuda/fence_cuda_test.cc
@@ -127,7 +127,8 @@ TEST(CUDAFenceTests, DISABLED_PartOnCPU) {
   session.Run(std::unordered_map<std::string, OrtValue>{{"X1", value}}, std::vector<std::string>{"Out"}, &outputs);
   ASSERT_TRUE(1 == outputs.size());
   const Tensor& output = outputs[0].Get<Tensor>();
-  EXPECT_EQ(output.Shape(), shape);
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&output.Shape()), *reinterpret_cast<const std::vector<int64_t>*>(&shape));
   EXPECT_EQ(output.DataType(), DataTypeImpl::GetType<float>());
 
   float expected_output[4] = {13.0f, -18.0f, -27.0f, 40.0f};
@@ -179,7 +180,8 @@ TEST(CUDAFenceTests, TileWithInitializer) {
   session.Run(std::unordered_map<std::string, OrtValue>{{"X1", value}}, std::vector<std::string>{"Y"}, &outputs);
   ASSERT_TRUE(1 == outputs.size());
   const Tensor& output = outputs[0].Get<Tensor>();
-  EXPECT_EQ(output.Shape(), TensorShape({2, 4}));
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&output.Shape()), (std::vector<int64_t>{2, 4}));
   EXPECT_EQ(output.DataType(), DataTypeImpl::GetType<float>());
 
   float expected_output[8] = {-1, 2, -1, 2, 3, -4, 3, -4};
@@ -242,7 +244,8 @@ TEST(CUDAFenceTests, TileWithComputedInput) {
   session.Run(std::unordered_map<std::string, OrtValue>{{"X1", value}}, std::vector<std::string>{"Out"}, &outputs);
   ASSERT_TRUE(1 == outputs.size());
   const Tensor& output = outputs[0].Get<Tensor>();
-  EXPECT_EQ(output.Shape(), TensorShape({4, 4}));
+  //Use reinterpret_cast to bypass a gcc bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51213
+  EXPECT_EQ(*reinterpret_cast<const std::vector<int64_t>*>(&output.Shape()), (std::vector<int64_t>{4, 4}));
   EXPECT_EQ(output.DataType(), DataTypeImpl::GetType<float>());
 
   float expected_output[16] = {7, -10, 7, -10, -15, 22, -15, 22, 7, -10, 7, -10, -15, 22, -15, 22};

From dcdebb409a1b652c7e308b38472b6705dd4de8c3 Mon Sep 17 00:00:00 2001
From: Ryan Lai <ryalai96@gmail.com>
Date: Wed, 15 Jan 2020 12:13:56 -0800
Subject: [PATCH 63/75] Refactor WinMLAPI Tests to build both google and taef
 test based on preprocessor definition (#2829)

* Add winml macro wrappers on top of google test macros

* change test methods to disabled

* Add custom winml macros for both taef and google tests

* PR comments

* Refactor winml api tests

* Move additional gtest specific macro definition into googleTestMacros.h
---
 cmake/winml_unittests.cmake                   |   1 +
 winml/test/api/APITest.h                      |  52 +--
 winml/test/api/LearningModelAPITest.cpp       | 369 +++++++++---------
 winml/test/api/LearningModelAPITest.h         |  41 ++
 .../test/api/LearningModelBindingAPITest.cpp  | 354 +++++++++--------
 winml/test/api/LearningModelBindingAPITest.h  |  49 +++
 .../test/api/LearningModelSessionAPITest.cpp  | 285 ++++++++------
 winml/test/api/LearningModelSessionAPITest.h  |  44 +++
 winml/test/common/googleTestMacros.h          |  19 +-
 winml/test/common/std.h                       |  17 +-
 winml/test/common/taefTestMacros.h            |   4 +-
 11 files changed, 723 insertions(+), 512 deletions(-)
 create mode 100644 winml/test/api/LearningModelAPITest.h
 create mode 100644 winml/test/api/LearningModelBindingAPITest.h
 create mode 100644 winml/test/api/LearningModelSessionAPITest.h

diff --git a/cmake/winml_unittests.cmake b/cmake/winml_unittests.cmake
index 95661e85d8bc2..4bdbb34432466 100644
--- a/cmake/winml_unittests.cmake
+++ b/cmake/winml_unittests.cmake
@@ -69,6 +69,7 @@ add_winml_test(
   SOURCES ${winml_test_api_src}
   LIBS winml_test_common
 )
+target_compile_definitions(winml_test_api PRIVATE BUILD_GOOGLE_TEST)
 target_precompiled_header(winml_test_api testPch.h)
 
 if (onnxruntime_USE_DML)
diff --git a/winml/test/api/APITest.h b/winml/test/api/APITest.h
index 68e68f724ddf5..fd40a19b787f4 100644
--- a/winml/test/api/APITest.h
+++ b/winml/test/api/APITest.h
@@ -5,37 +5,25 @@
 //-----------------------------------------------------------------------------
 
 #pragma once
+#include "fileHelpers.h"
+namespace APITest {
+static void LoadModel(const std::wstring& modelPath,
+                      winrt::Windows::AI::MachineLearning::LearningModel& learningModel) {
+  std::wstring fullPath = FileHelpers::GetModulePath() + modelPath;
+  learningModel = winrt::Windows::AI::MachineLearning::LearningModel::LoadFromFilePath(fullPath);
+};
 
-#include <gtest/gtest.h>
-
-class APITest : public ::testing::Test
-{
-protected:
-    void LoadModel(const std::wstring& modelPath)
-    {
-        std::wstring fullPath = FileHelpers::GetModulePath() + modelPath;
-        m_model = winrt::Windows::AI::MachineLearning::LearningModel::LoadFromFilePath(fullPath);
-    }
-
-    winrt::Windows::AI::MachineLearning::LearningModel m_model = nullptr;
-    winrt::Windows::AI::MachineLearning::LearningModelDevice m_device = nullptr;
-    winrt::Windows::AI::MachineLearning::LearningModelSession m_session = nullptr;
-
-    uint64_t GetAdapterIdQuadPart()
-    {
-        LARGE_INTEGER id;
-        id.LowPart = m_device.AdapterId().LowPart;
-        id.HighPart = m_device.AdapterId().HighPart;
-        return id.QuadPart;
-    };
-
-    _LUID GetAdapterIdAsLUID()
-    {
-        _LUID id;
-        id.LowPart = m_device.AdapterId().LowPart;
-        id.HighPart = m_device.AdapterId().HighPart;
-        return id;
-    }
-
-    bool m_runGPUTests = true;
+static uint64_t GetAdapterIdQuadPart(winrt::Windows::AI::MachineLearning::LearningModelDevice& device) {
+  LARGE_INTEGER id;
+  id.LowPart = device.AdapterId().LowPart;
+  id.HighPart = device.AdapterId().HighPart;
+  return id.QuadPart;
 };
+
+static _LUID GetAdapterIdAsLUID(winrt::Windows::AI::MachineLearning::LearningModelDevice& device) {
+  _LUID id;
+  id.LowPart = device.AdapterId().LowPart;
+  id.HighPart = device.AdapterId().HighPart;
+  return id;
+}
+};  // namespace APITest
diff --git a/winml/test/api/LearningModelAPITest.cpp b/winml/test/api/LearningModelAPITest.cpp
index ad592a9a86301..639940551d580 100644
--- a/winml/test/api/LearningModelAPITest.cpp
+++ b/winml/test/api/LearningModelAPITest.cpp
@@ -1,7 +1,6 @@
 #include "testPch.h"
-
+#include "LearningModelAPITest.h"
 #include "APITest.h"
-
 #include <winrt/Windows.Graphics.Imaging.h>
 #include <winrt/Windows.Media.h>
 #include <winrt/Windows.Storage.h>
@@ -15,107 +14,96 @@ using namespace winrt::Windows::Media;
 using namespace winrt::Windows::Storage;
 using namespace winrt::Windows::Storage::Streams;
 
-class LearningModelAPITest : public APITest
-{
-protected:
-    LearningModelAPITest() {
-        init_apartment();
-        m_model = nullptr;
-        m_device = nullptr;
-        m_session = nullptr;
-    }
-};
-
-class LearningModelAPITestGpu : public LearningModelAPITest
-{
-protected:
-    void SetUp() override
-    {
-        GPUTEST
-    }
-};
+static void LearningModelAPITestSetup() {
+  init_apartment();
+}
 
-TEST_F(LearningModelAPITest, CreateModelFromFilePath)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
+static void LearningModelAPITestGpuSetup() {
+  GPUTEST;
+  init_apartment();
 }
 
-TEST_F(LearningModelAPITest, CreateModelFromIStorage)
-{
-    std::wstring path = FileHelpers::GetModulePath() + L"squeezenet_modifiedforruntimestests.onnx";
-    auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromStorageFileAsync(storageFile).get());
-    EXPECT_TRUE(m_model != nullptr);
-
-    // check the author so we know the model was populated correctly.
-    std::wstring author(m_model.Author());
-    EXPECT_EQ(L"onnx-caffe2", author);
+static void CreateModelFromFilePath() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
 }
 
-TEST_F(LearningModelAPITest, CreateModelFromIStorageOutsideCwd)
-{
-    std::wstring path = FileHelpers::GetModulePath() + L"ModelSubdirectory\\ModelInSubdirectory.onnx";
-    auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromStorageFileAsync(storageFile).get());
-    EXPECT_TRUE(m_model != nullptr);
-
-    // check the author so we know the model was populated correctly.
-    std::wstring author(m_model.Author());
-    EXPECT_EQ(L"onnx-caffe2", author);
+static void CreateModelFromIStorage() {
+  std::wstring path = FileHelpers::GetModulePath() + L"squeezenet_modifiedforruntimestests.onnx";
+  auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromStorageFileAsync(storageFile).get());
+  WINML_EXPECT_TRUE(learningModel != nullptr);
+
+  // check the author so we know the model was populated correctly.
+  std::wstring author(learningModel.Author());
+  WINML_EXPECT_EQUAL(L"onnx-caffe2", author);
 }
 
-TEST_F(LearningModelAPITest, CreateModelFromIStream)
-{
-    std::wstring path = FileHelpers::GetModulePath() + L"squeezenet_modifiedforruntimestests.onnx";
-    auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
-    winrt::Windows::Storage::Streams::IRandomAccessStreamReference streamref;
-    storageFile.as(streamref);
+static void CreateModelFromIStorageOutsideCwd() {
+  std::wstring path = FileHelpers::GetModulePath() + L"ModelSubdirectory\\ModelInSubdirectory.onnx";
+  auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromStorageFileAsync(storageFile).get());
+  WINML_EXPECT_TRUE(learningModel != nullptr);
 
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromStreamAsync(streamref).get());
-    EXPECT_TRUE(m_model != nullptr);
+  // check the author so we know the model was populated correctly.
+  std::wstring author(learningModel.Author());
+  WINML_EXPECT_EQUAL(L"onnx-caffe2", author);
+}
 
-    // check the author so we know the model was populated correctly.
-    std::wstring author(m_model.Author());
-    EXPECT_EQ(L"onnx-caffe2", author);
+static void CreateModelFromIStream() {
+  std::wstring path = FileHelpers::GetModulePath() + L"squeezenet_modifiedforruntimestests.onnx";
+  auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
+  winrt::Windows::Storage::Streams::IRandomAccessStreamReference streamref;
+  storageFile.as(streamref);
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromStreamAsync(streamref).get());
+  WINML_EXPECT_TRUE(learningModel != nullptr);
+
+  // check the author so we know the model was populated correctly.
+  std::wstring author(learningModel.Author());
+  WINML_EXPECT_EQUAL(L"onnx-caffe2", author);
 }
 
-TEST_F(LearningModelAPITest, GetAuthor)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
-    std::wstring author(m_model.Author());
-    EXPECT_EQ(L"onnx-caffe2", author);
+static void ModelGetAuthor() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
+  std::wstring author(learningModel.Author());
+  WINML_EXPECT_EQUAL(L"onnx-caffe2", author);
 }
 
-TEST_F(LearningModelAPITest, GetName)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
-    std::wstring name(m_model.Name());
-    EXPECT_EQ(L"squeezenet_old", name);
+static void ModelGetName() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
+  std::wstring name(learningModel.Name());
+  WINML_EXPECT_EQUAL(L"squeezenet_old", name);
 }
 
-TEST_F(LearningModelAPITest, GetDomain)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
-    std::wstring domain(m_model.Domain());
-    EXPECT_EQ(L"test-domain", domain);
+static void ModelGetDomain() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
+  std::wstring domain(learningModel.Domain());
+  WINML_EXPECT_EQUAL(L"test-domain", domain);
 }
 
-TEST_F(LearningModelAPITest, GetDescription)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
-    std::wstring description(m_model.Description());
-    EXPECT_EQ(L"test-doc_string", description);
+static void ModelGetDescription() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
+  std::wstring description(learningModel.Description());
+  WINML_EXPECT_EQUAL(L"test-doc_string", description);
 }
 
-TEST_F(LearningModelAPITest, GetVersion)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
-    int64_t version(m_model.Version());
-    (void)(version);
+static void ModelGetVersion() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
+  int64_t version(learningModel.Version());
+  (void)(version);
 }
 
 typedef std::vector<std::pair<std::wstring, std::wstring>> Metadata;
 
+/*
 class MetadataTest : public LearningModelAPITest, public testing::WithParamInterface<std::pair<std::wstring, Metadata>>
 {};
 
@@ -125,16 +113,16 @@ TEST_P(MetadataTest, GetMetaData)
     std::vector<std::pair<std::wstring, std::wstring>> keyValuePairs;
 
     tie(fileName, keyValuePairs) = GetParam();
-    EXPECT_NO_THROW(LoadModel(fileName.c_str()));
-    EXPECT_TRUE(m_model.Metadata() != nullptr);
-    EXPECT_EQ(keyValuePairs.size(), m_model.Metadata().Size());
+    WINML_EXPECT_NO_THROW(LoadModel(fileName.c_str()));
+    WINML_EXPECT_TRUE(m_model.Metadata() != nullptr);
+    WINML_EXPECT_EQUAL(keyValuePairs.size(), m_model.Metadata().Size());
 
     auto iter = m_model.Metadata().First();
     for (auto& keyValue : keyValuePairs)
     {
-        EXPECT_TRUE(iter.HasCurrent());
-        EXPECT_EQ(keyValue.first, std::wstring(iter.Current().Key()));
-        EXPECT_EQ(keyValue.second, std::wstring(iter.Current().Value()));
+        WINML_EXPECT_TRUE(iter.HasCurrent());
+        WINML_EXPECT_EQUAL(keyValue.first, std::wstring(iter.Current().Key()));
+        WINML_EXPECT_EQUAL(keyValue.second, std::wstring(iter.Current().Value()));
         iter.MoveNext();
     }
 }
@@ -147,122 +135,141 @@ INSTANTIATE_TEST_SUITE_P(
         std::pair(L"modelWithMetaData.onnx", Metadata{{L"thisisalongkey", L"thisisalongvalue"}}),
         std::pair(L"modelWith2MetaData.onnx", Metadata{{L"thisisalongkey", L"thisisalongvalue"}, {L"key2", L"val2"}})
 ));
+*/
 
-TEST_F(LearningModelAPITest, EnumerateInputs)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
+static void EnumerateInputs() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
 
-    // purposely don't cache "InputFeatures" in order to exercise calling it multiple times
-    EXPECT_TRUE(m_model.InputFeatures().First().HasCurrent());
+  // purposely don't cache "InputFeatures" in order to exercise calling it multiple times
+  WINML_EXPECT_TRUE(learningModel.InputFeatures().First().HasCurrent());
 
-    std::wstring name(m_model.InputFeatures().First().Current().Name());
-    EXPECT_EQ(L"data_0", name);
+  std::wstring name(learningModel.InputFeatures().First().Current().Name());
+  WINML_EXPECT_EQUAL(L"data_0", name);
 
-    // make sure it's either tensor or image
-    TensorFeatureDescriptor tensorDescriptor = nullptr;
-    m_model.InputFeatures().First().Current().try_as(tensorDescriptor);
-    if (tensorDescriptor == nullptr)
-    {
-        ImageFeatureDescriptor imageDescriptor = nullptr;
-        EXPECT_NO_THROW(m_model.InputFeatures().First().Current().as(imageDescriptor));
-    }
+  // make sure it's either tensor or image
+  TensorFeatureDescriptor tensorDescriptor = nullptr;
+  learningModel.InputFeatures().First().Current().try_as(tensorDescriptor);
+  if (tensorDescriptor == nullptr) {
+    ImageFeatureDescriptor imageDescriptor = nullptr;
+    WINML_EXPECT_NO_THROW(learningModel.InputFeatures().First().Current().as(imageDescriptor));
+  }
 
-    auto modelDataKind = tensorDescriptor.TensorKind();
-    EXPECT_EQ(TensorKind::Float, modelDataKind);
+  auto modelDataKind = tensorDescriptor.TensorKind();
+  WINML_EXPECT_EQUAL(TensorKind::Float, modelDataKind);
 
-    EXPECT_TRUE(tensorDescriptor.IsRequired());
+  WINML_EXPECT_TRUE(tensorDescriptor.IsRequired());
 
-    std::vector<int64_t> expectedShapes = { 1,3,224,224 };
-    EXPECT_EQ(expectedShapes.size(), tensorDescriptor.Shape().Size());
-    for (uint32_t j = 0; j < tensorDescriptor.Shape().Size(); j++)
-    {
-        EXPECT_EQ(expectedShapes.at(j), tensorDescriptor.Shape().GetAt(j));
-    }
+  std::vector<int64_t> expectedShapes = {1, 3, 224, 224};
+  WINML_EXPECT_EQUAL(expectedShapes.size(), tensorDescriptor.Shape().Size());
+  for (uint32_t j = 0; j < tensorDescriptor.Shape().Size(); j++) {
+    WINML_EXPECT_EQUAL(expectedShapes.at(j), tensorDescriptor.Shape().GetAt(j));
+  }
 
-    auto first = m_model.InputFeatures().First();
-    first.MoveNext();
-    EXPECT_FALSE(first.HasCurrent());
+  auto first = learningModel.InputFeatures().First();
+  first.MoveNext();
+  WINML_EXPECT_FALSE(first.HasCurrent());
 }
 
-TEST_F(LearningModelAPITest, EnumerateOutputs)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
+static void EnumerateOutputs() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
 
-    // purposely don't cache "OutputFeatures" in order to exercise calling it multiple times
-    std::wstring name(m_model.OutputFeatures().First().Current().Name());
-    EXPECT_EQ(L"softmaxout_1", name);
+  // purposely don't cache "OutputFeatures" in order to exercise calling it multiple times
+  std::wstring name(learningModel.OutputFeatures().First().Current().Name());
+  WINML_EXPECT_EQUAL(L"softmaxout_1", name);
 
-    TensorFeatureDescriptor tensorDescriptor = nullptr;
-    EXPECT_NO_THROW(m_model.OutputFeatures().First().Current().as(tensorDescriptor));
-    EXPECT_TRUE(tensorDescriptor != nullptr);
+  TensorFeatureDescriptor tensorDescriptor = nullptr;
+  WINML_EXPECT_NO_THROW(learningModel.OutputFeatures().First().Current().as(tensorDescriptor));
+  WINML_EXPECT_TRUE(tensorDescriptor != nullptr);
 
-    auto tensorName = tensorDescriptor.Name();
-    EXPECT_EQ(L"softmaxout_1", tensorName);
+  auto tensorName = tensorDescriptor.Name();
+  WINML_EXPECT_EQUAL(L"softmaxout_1", tensorName);
 
-    auto modelDataKind = tensorDescriptor.TensorKind();
-    EXPECT_EQ(TensorKind::Float, modelDataKind);
+  auto modelDataKind = tensorDescriptor.TensorKind();
+  WINML_EXPECT_EQUAL(TensorKind::Float, modelDataKind);
 
-    EXPECT_TRUE(tensorDescriptor.IsRequired());
+  WINML_EXPECT_TRUE(tensorDescriptor.IsRequired());
 
-    std::vector<int64_t> expectedShapes = { 1, 1000, 1, 1 };
-    EXPECT_EQ(expectedShapes.size(), tensorDescriptor.Shape().Size());
-    for (uint32_t j = 0; j < tensorDescriptor.Shape().Size(); j++)
-    {
-        EXPECT_EQ(expectedShapes.at(j), tensorDescriptor.Shape().GetAt(j));
-    }
+  std::vector<int64_t> expectedShapes = {1, 1000, 1, 1};
+  WINML_EXPECT_EQUAL(expectedShapes.size(), tensorDescriptor.Shape().Size());
+  for (uint32_t j = 0; j < tensorDescriptor.Shape().Size(); j++) {
+    WINML_EXPECT_EQUAL(expectedShapes.at(j), tensorDescriptor.Shape().GetAt(j));
+  }
 
-    auto first = m_model.OutputFeatures().First();
-    first.MoveNext();
-    EXPECT_FALSE(first.HasCurrent());
+  auto first = learningModel.OutputFeatures().First();
+  first.MoveNext();
+  WINML_EXPECT_FALSE(first.HasCurrent());
 }
 
-TEST_F(LearningModelAPITest, CloseModelCheckMetadata)
-{
-    EXPECT_NO_THROW(LoadModel(L"squeezenet_modifiedforruntimestests.onnx"));
-    EXPECT_NO_THROW(m_model.Close());
-    std::wstring author(m_model.Author());
-    EXPECT_EQ(L"onnx-caffe2", author);
-    std::wstring name(m_model.Name());
-    EXPECT_EQ(L"squeezenet_old", name);
-    std::wstring domain(m_model.Domain());
-    EXPECT_EQ(L"test-domain", domain);
-    std::wstring description(m_model.Description());
-    EXPECT_EQ(L"test-doc_string", description);
-    int64_t version(m_model.Version());
-    EXPECT_EQ(123456, version);
+static void CloseModelCheckMetadata() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"squeezenet_modifiedforruntimestests.onnx", learningModel));
+  WINML_EXPECT_NO_THROW(learningModel.Close());
+  std::wstring author(learningModel.Author());
+  WINML_EXPECT_EQUAL(L"onnx-caffe2", author);
+  std::wstring name(learningModel.Name());
+  WINML_EXPECT_EQUAL(L"squeezenet_old", name);
+  std::wstring domain(learningModel.Domain());
+  WINML_EXPECT_EQUAL(L"test-domain", domain);
+  std::wstring description(learningModel.Description());
+  WINML_EXPECT_EQUAL(L"test-doc_string", description);
+  int64_t version(learningModel.Version());
+  WINML_EXPECT_EQUAL(123456, version);
 }
 
-TEST_F(LearningModelAPITestGpu, CloseModelCheckEval)
-{
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
-    LearningModelSession session = nullptr;
-    EXPECT_NO_THROW(session = LearningModelSession(m_model));
-    EXPECT_NO_THROW(m_model.Close());
-
-    std::wstring fullImagePath = FileHelpers::GetModulePath() + L"kitten_224.png";
-    StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
-    IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
-    SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
-    VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
-
-    LearningModelBinding binding = nullptr;
-    EXPECT_NO_THROW(binding = LearningModelBinding(session));
-    EXPECT_NO_THROW(binding.Bind(m_model.InputFeatures().First().Current().Name(), frame));
-
-    EXPECT_NO_THROW(session.Evaluate(binding, L""));
+static void CloseModelCheckEval() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
+  LearningModelSession session = nullptr;
+  WINML_EXPECT_NO_THROW(session = LearningModelSession(learningModel));
+  WINML_EXPECT_NO_THROW(learningModel.Close());
+
+  std::wstring fullImagePath = FileHelpers::GetModulePath() + L"kitten_224.png";
+  StorageFile imagefile = StorageFile::GetFileFromPathAsync(fullImagePath).get();
+  IRandomAccessStream stream = imagefile.OpenAsync(FileAccessMode::Read).get();
+  SoftwareBitmap softwareBitmap = (BitmapDecoder::CreateAsync(stream).get()).GetSoftwareBitmapAsync().get();
+  VideoFrame frame = VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
+
+  LearningModelBinding binding = nullptr;
+  WINML_EXPECT_NO_THROW(binding = LearningModelBinding(session));
+  WINML_EXPECT_NO_THROW(binding.Bind(learningModel.InputFeatures().First().Current().Name(), frame));
+
+  WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
-TEST_F(LearningModelAPITest, CloseModelNoNewSessions)
-{
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
-    EXPECT_NO_THROW(m_model.Close());
-    LearningModelSession session = nullptr;
-    EXPECT_THROW(
-        try {
-            session = LearningModelSession(m_model);
-        } catch (const winrt::hresult_error& e) {
-            EXPECT_EQ(E_INVALIDARG, e.code());
-            throw;
-        }
-    , winrt::hresult_error);
+static void CloseModelNoNewSessions() {
+  LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
+  WINML_EXPECT_NO_THROW(learningModel.Close());
+  LearningModelSession session = nullptr;
+  WINML_EXPECT_THROW_SPECIFIC(
+      session = LearningModelSession(learningModel);,
+      winrt::hresult_error,
+      [](const winrt::hresult_error& e) -> bool {
+            return e.code() == E_INVALIDARG;
+      });
 }
+
+const LearningModelApiTestApi& getapi() {
+  static constexpr LearningModelApiTestApi api =
+  {
+    LearningModelAPITestSetup,
+    LearningModelAPITestGpuSetup,
+    CreateModelFromFilePath,
+    CreateModelFromIStorage,
+    CreateModelFromIStorageOutsideCwd,
+    CreateModelFromIStream,
+    ModelGetAuthor,
+    ModelGetName,
+    ModelGetDomain,
+    ModelGetDescription,
+    ModelGetVersion,
+    EnumerateInputs,
+    EnumerateOutputs,
+    CloseModelCheckMetadata,
+    CloseModelCheckEval,
+    CloseModelNoNewSessions
+  };
+  return api;
+}
\ No newline at end of file
diff --git a/winml/test/api/LearningModelAPITest.h b/winml/test/api/LearningModelAPITest.h
new file mode 100644
index 0000000000000..4a723ab2d9d29
--- /dev/null
+++ b/winml/test/api/LearningModelAPITest.h
@@ -0,0 +1,41 @@
+#include "test.h"
+struct LearningModelApiTestApi
+{
+  SetupTest LearningModelAPITestSetup;
+  SetupTest LearningModelAPITestGpuSetup;
+  VoidTest CreateModelFromFilePath;
+  VoidTest CreateModelFromIStorage;
+  VoidTest CreateModelFromIStorageOutsideCwd;
+  VoidTest CreateModelFromIStream;
+  VoidTest ModelGetAuthor;
+  VoidTest ModelGetName;
+  VoidTest ModelGetDomain;
+  VoidTest ModelGetDescription;
+  VoidTest ModelGetVersion;
+  VoidTest EnumerateInputs;
+  VoidTest EnumerateOutputs;
+  VoidTest CloseModelCheckMetadata;
+  VoidTest CloseModelCheckEval;
+  VoidTest CloseModelNoNewSessions;
+};
+const LearningModelApiTestApi& getapi();
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelAPITest, LearningModelAPITestSetup)
+WINML_TEST(LearningModelAPITest, CreateModelFromFilePath)
+WINML_TEST(LearningModelAPITest, CreateModelFromIStorage)
+WINML_TEST(LearningModelAPITest, CreateModelFromIStorageOutsideCwd)
+WINML_TEST(LearningModelAPITest, CreateModelFromIStream)
+WINML_TEST(LearningModelAPITest, ModelGetAuthor)
+WINML_TEST(LearningModelAPITest, ModelGetName)
+WINML_TEST(LearningModelAPITest, ModelGetDomain)
+WINML_TEST(LearningModelAPITest, ModelGetDescription)
+WINML_TEST(LearningModelAPITest, ModelGetVersion)
+WINML_TEST(LearningModelAPITest, EnumerateInputs)
+WINML_TEST(LearningModelAPITest, EnumerateOutputs)
+WINML_TEST(LearningModelAPITest, CloseModelCheckMetadata)
+WINML_TEST(LearningModelAPITest, CloseModelNoNewSessions)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelAPITestGpu, LearningModelAPITestGpuSetup)
+WINML_TEST(LearningModelAPITestGpu, CloseModelCheckEval)
+WINML_TEST_CLASS_END()
\ No newline at end of file
diff --git a/winml/test/api/LearningModelBindingAPITest.cpp b/winml/test/api/LearningModelBindingAPITest.cpp
index 969c28ba63937..99d9dc6e9b251 100644
--- a/winml/test/api/LearningModelBindingAPITest.cpp
+++ b/winml/test/api/LearningModelBindingAPITest.cpp
@@ -1,13 +1,14 @@
 #include "testPch.h"
 
 #include "APITest.h"
+#include "LearningModelBindingAPITest.h"
 #include "SqueezeNetValidator.h"
 
 #include <winrt/Windows.Graphics.Imaging.h>
 #include <winrt/Windows.Media.h>
 #include "winrt/Windows.Storage.h"
 #include "DeviceHelpers.h"
-
+#include <sstream>
 using namespace winrt;
 using namespace winrt::Windows::AI::MachineLearning;
 using namespace winrt::Windows::Foundation::Collections;
@@ -15,25 +16,22 @@ using namespace winrt::Windows::Graphics::Imaging;
 using namespace winrt::Windows::Media;
 using namespace winrt::Windows::Storage;
 
-class LearningModelBindingAPITest : public APITest
-{};
+static void LearningModelBindingAPITestSetup() {
+  init_apartment();
+}
 
-class LearningModelBindingAPITestGpu : public LearningModelBindingAPITest
-{
-protected:
-    void SetUp() override
-    {
-        GPUTEST
-    }
-};
+static void LearningModelBindingAPITestGpuSetup() {
+  GPUTEST;
+  init_apartment();
+}
 
-TEST_F(LearningModelBindingAPITest, CpuSqueezeNet)
+static void CpuSqueezeNet()
 {
     std::string cpuInstance("CPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(cpuInstance, LearningModelDeviceKind::Cpu, /*dataTolerance*/ 0.00001f, false);
 }
 
-TEST_F(LearningModelBindingAPITest, CpuSqueezeNetEmptyOutputs)
+static void CpuSqueezeNetEmptyOutputs()
 {
     std::string cpuInstance("CPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(
@@ -44,7 +42,7 @@ TEST_F(LearningModelBindingAPITest, CpuSqueezeNetEmptyOutputs)
         OutputBindingStrategy::Empty);
 }
 
-TEST_F(LearningModelBindingAPITest, CpuSqueezeNetUnboundOutputs)
+static void CpuSqueezeNetUnboundOutputs()
 {
     std::string cpuInstance("CPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(
@@ -55,7 +53,7 @@ TEST_F(LearningModelBindingAPITest, CpuSqueezeNetUnboundOutputs)
         OutputBindingStrategy::Unbound);
 }
 
-TEST_F(LearningModelBindingAPITest, CpuSqueezeNetBindInputTensorAsInspectable)
+static void CpuSqueezeNetBindInputTensorAsInspectable()
 {
     std::string cpuInstance("CPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(
@@ -67,27 +65,28 @@ TEST_F(LearningModelBindingAPITest, CpuSqueezeNetBindInputTensorAsInspectable)
         true /* bind inputs as inspectables */);
 }
 
-TEST_F(LearningModelBindingAPITest, CastMapInt64)
+static void CastMapInt64()
 {
-    EXPECT_NO_THROW(LoadModel(L"castmap-int64.onnx"));
+    WINML_EXPECT_NO_THROW(LearningModel::LoadFromFilePath(FileHelpers::GetModulePath() + L"castmap-int64.onnx"));
     // TODO: Check Descriptor
 }
 
-TEST_F(LearningModelBindingAPITest, DictionaryVectorizerMapInt64)
+static void DictionaryVectorizerMapInt64()
 {
-    EXPECT_NO_THROW(LoadModel(L"dictvectorizer-int64.onnx"));
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"dictvectorizer-int64.onnx", learningModel));
 
-    auto inputDescriptor = m_model.InputFeatures().First().Current();
-    EXPECT_TRUE(inputDescriptor.Kind() == LearningModelFeatureKind::Map);
+    auto inputDescriptor = learningModel.InputFeatures().First().Current();
+    WINML_EXPECT_TRUE(inputDescriptor.Kind() == LearningModelFeatureKind::Map);
     auto mapDescriptor = inputDescriptor.as<MapFeatureDescriptor>();
-    EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::Int64);
-    EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
+    WINML_EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::Int64);
+    WINML_EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
     auto tensorDescriptor = mapDescriptor.ValueDescriptor().as<TensorFeatureDescriptor>();
     // empty size means tensor of scalar value
-    EXPECT_TRUE(tensorDescriptor.Shape().Size() == 0);
-    EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
+    WINML_EXPECT_TRUE(tensorDescriptor.Shape().Size() == 0);
+    WINML_EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
 
-    LearningModelSession modelSession(m_model);
+    LearningModelSession modelSession(learningModel);
     LearningModelBinding binding(modelSession);
     std::unordered_map<int64_t, float> map;
     map[1] = 1.f;
@@ -102,38 +101,39 @@ TEST_F(LearningModelBindingAPITest, DictionaryVectorizerMapInt64)
     binding.Bind(mapInputName, abiMap);
     auto mapInputInspectable = abiMap.as<winrt::Windows::Foundation::IInspectable>();
     auto first = binding.First();
-    EXPECT_TRUE(first.Current().Key() == mapInputName);
-    EXPECT_TRUE(first.Current().Value() == mapInputInspectable);
-    EXPECT_TRUE(binding.Lookup(mapInputName) == mapInputInspectable);
+    WINML_EXPECT_TRUE(first.Current().Key() == mapInputName);
+    WINML_EXPECT_TRUE(first.Current().Value() == mapInputInspectable);
+    WINML_EXPECT_TRUE(binding.Lookup(mapInputName) == mapInputInspectable);
 
     // Bind as IMapView
     auto mapView = abiMap.GetView();
     binding.Bind(mapInputName, mapView);
     mapInputInspectable = mapView.as<winrt::Windows::Foundation::IInspectable>();
     first = binding.First();
-    EXPECT_TRUE(first.Current().Key() == mapInputName);
-    EXPECT_TRUE(first.Current().Value() == mapView);
-    EXPECT_TRUE(binding.Lookup(mapInputName) == mapView);
+    WINML_EXPECT_TRUE(first.Current().Key() == mapInputName);
+    WINML_EXPECT_TRUE(first.Current().Value() == mapView);
+    WINML_EXPECT_TRUE(binding.Lookup(mapInputName) == mapView);
 
 }
 
-TEST_F(LearningModelBindingAPITest, DictionaryVectorizerMapString)
+static void DictionaryVectorizerMapString()
 {
-    EXPECT_NO_THROW(LoadModel(L"dictvectorizer-string.onnx"));
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"dictvectorizer-string.onnx", learningModel));
 
-    auto inputDescriptor = m_model.InputFeatures().First().Current();
-    EXPECT_TRUE(inputDescriptor.Kind() == LearningModelFeatureKind::Map);
+    auto inputDescriptor = learningModel.InputFeatures().First().Current();
+    WINML_EXPECT_TRUE(inputDescriptor.Kind() == LearningModelFeatureKind::Map);
 
     auto mapDescriptor = inputDescriptor.as<MapFeatureDescriptor>();
-    EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::String);
-    EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
+    WINML_EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::String);
+    WINML_EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
 
     auto tensorDescriptor = mapDescriptor.ValueDescriptor().as<TensorFeatureDescriptor>();
     // empty size means tensor of scalar value
-    EXPECT_TRUE(tensorDescriptor.Shape().Size() == 0);
-    EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
+    WINML_EXPECT_TRUE(tensorDescriptor.Shape().Size() == 0);
+    WINML_EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
 
-    LearningModelSession modelSession(m_model);
+    LearningModelSession modelSession(learningModel);
     LearningModelBinding binding(modelSession);
     std::unordered_map<winrt::hstring, float> map;
     map[L"1"] = 1.f;
@@ -146,9 +146,9 @@ TEST_F(LearningModelBindingAPITest, DictionaryVectorizerMapString)
 
     auto mapInputInspectable = abiMap.as<winrt::Windows::Foundation::IInspectable>();
     auto first = binding.First();
-    EXPECT_TRUE(first.Current().Key() == mapInputName);
-    EXPECT_TRUE(first.Current().Value() == mapInputInspectable);
-    EXPECT_TRUE(binding.Lookup(mapInputName) == mapInputInspectable);
+    WINML_EXPECT_TRUE(first.Current().Key() == mapInputName);
+    WINML_EXPECT_TRUE(first.Current().Value() == mapInputInspectable);
+    WINML_EXPECT_TRUE(binding.Lookup(mapInputName) == mapInputInspectable);
 }
 
 static void RunZipMapInt64(
@@ -157,15 +157,15 @@ static void RunZipMapInt64(
 {
     auto outputFeatures = model.OutputFeatures();
     auto outputDescriptor = outputFeatures.First().Current();
-    EXPECT_TRUE(outputDescriptor.Kind() == LearningModelFeatureKind::Sequence);
+    WINML_EXPECT_TRUE(outputDescriptor.Kind() == LearningModelFeatureKind::Sequence);
 
     auto seqDescriptor = outputDescriptor.as<SequenceFeatureDescriptor>();
     auto mapDescriptor = seqDescriptor.ElementDescriptor().as<MapFeatureDescriptor>();
-    EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::Int64);
+    WINML_EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::Int64);
 
-    EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
+    WINML_EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
     auto tensorDescriptor = mapDescriptor.ValueDescriptor().as<TensorFeatureDescriptor>();
-    EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
+    WINML_EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
 
     LearningModelSession session(model);
     LearningModelBinding binding(session);
@@ -199,63 +199,66 @@ static void RunZipMapInt64(
         // from output binding
         const auto &out1 = abiOutput.GetAt(0);
         const auto &out2 = result.Lookup(L"Y").as<IVectorView<ABIMap>>().GetAt(0);
-        SCOPED_TRACE((std::ostringstream() << "size: " << out1.Size()).str());
+        WINML_LOG_COMMENT((std::ostringstream() << "size: " << out1.Size()).str());
         // check outputs
         auto iter1 = out1.First();
         auto iter2 = out2.First();
         for (uint32_t i = 0, size = (uint32_t)inputs.size(); i < size; ++i)
         {
-            EXPECT_TRUE(iter1.HasCurrent());
-            EXPECT_TRUE(iter2.HasCurrent());
+            WINML_EXPECT_TRUE(iter1.HasCurrent());
+            WINML_EXPECT_TRUE(iter2.HasCurrent());
             const auto &pair1 = iter1.Current();
             const auto &pair2 = iter2.Current();
-            SCOPED_TRACE((std::ostringstream() << "key: " << pair1.Key() << ", value: " << pair2.Value()).str());
-            EXPECT_TRUE(pair1.Key() == i && pair2.Key() == i);
-            EXPECT_TRUE(pair1.Value() == inputs[i] && pair2.Value() == inputs[i]);
+            WINML_LOG_COMMENT((std::ostringstream() << "key: " << pair1.Key() << ", value: " << pair2.Value()).str());
+            WINML_EXPECT_TRUE(pair1.Key() == i && pair2.Key() == i);
+            WINML_EXPECT_TRUE(pair1.Value() == inputs[i] && pair2.Value() == inputs[i]);
             iter1.MoveNext();
             iter2.MoveNext();
         }
-        EXPECT_TRUE(!iter1.HasCurrent());
-        EXPECT_TRUE(!iter2.HasCurrent());
+        WINML_EXPECT_TRUE(!iter1.HasCurrent());
+        WINML_EXPECT_TRUE(!iter2.HasCurrent());
     }
     else
     {
         abiOutput = result.Lookup(L"Y").as<ABISequeneceOfMap>();
-        EXPECT_TRUE(abiOutput.Size() == 1);
+        WINML_EXPECT_TRUE(abiOutput.Size() == 1);
         ABIMap map = abiOutput.GetAt(0);
-        EXPECT_TRUE(map.Size() == 3);
-        EXPECT_TRUE(map.Lookup(0) == 0.5);
-        EXPECT_TRUE(map.Lookup(1) == .25);
-        EXPECT_TRUE(map.Lookup(2) == .125);
+        WINML_EXPECT_TRUE(map.Size() == 3);
+        WINML_EXPECT_TRUE(map.Lookup(0) == 0.5);
+        WINML_EXPECT_TRUE(map.Lookup(1) == .25);
+        WINML_EXPECT_TRUE(map.Lookup(2) == .125);
     }
 }
 
-TEST_F(LearningModelBindingAPITest, ZipMapInt64)
+static void ZipMapInt64()
 {
-    EXPECT_NO_THROW(LoadModel(L"zipmap-int64.onnx"));
-    RunZipMapInt64(m_model, OutputBindingStrategy::Bound);
+    LearningModel learningModel= nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"zipmap-int64.onnx", learningModel));
+    RunZipMapInt64(learningModel, OutputBindingStrategy::Bound);
 }
 
-TEST_F(LearningModelBindingAPITest, ZipMapInt64Unbound)
+static void ZipMapInt64Unbound()
 {
-    EXPECT_NO_THROW(LoadModel(L"zipmap-int64.onnx"));
-    RunZipMapInt64(m_model, OutputBindingStrategy::Unbound);
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"zipmap-int64.onnx", learningModel));
+    RunZipMapInt64(learningModel, OutputBindingStrategy::Unbound);
 }
 
-TEST_F(LearningModelBindingAPITest, ZipMapString)
+static void ZipMapString()
 {
     // output constraint: "seq(map(string, float))" or "seq(map(int64, float))"
-    EXPECT_NO_THROW(LoadModel(L"zipmap-string.onnx"));
-    auto outputs = m_model.OutputFeatures();
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"zipmap-string.onnx", learningModel));
+    auto outputs = learningModel.OutputFeatures();
     auto outputDescriptor = outputs.First().Current();
-    EXPECT_TRUE(outputDescriptor.Kind() == LearningModelFeatureKind::Sequence);
+    WINML_EXPECT_TRUE(outputDescriptor.Kind() == LearningModelFeatureKind::Sequence);
     auto mapDescriptor = outputDescriptor.as<SequenceFeatureDescriptor>().ElementDescriptor().as<MapFeatureDescriptor>();
-    EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::String);
-    EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
+    WINML_EXPECT_TRUE(mapDescriptor.KeyKind() == TensorKind::String);
+    WINML_EXPECT_TRUE(mapDescriptor.ValueDescriptor().Kind() == LearningModelFeatureKind::Tensor);
     auto tensorDescriptor = mapDescriptor.ValueDescriptor().as<TensorFeatureDescriptor>();
-    EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
+    WINML_EXPECT_TRUE(tensorDescriptor.TensorKind() == TensorKind::Float);
 
-    LearningModelSession session(m_model);
+    LearningModelSession session(learningModel);
     LearningModelBinding binding(session);
 
     std::vector<float> inputs = { 0.5f, 0.25f, 0.125f };
@@ -274,27 +277,27 @@ TEST_F(LearningModelBindingAPITest, ZipMapString)
     // from output binding
     const auto &out1 = ABIOutput.GetAt(0);
     const auto &out2 = result.Lookup(L"Y").as<IVectorView<ABIMap>>().GetAt(0);
-    SCOPED_TRACE((std::ostringstream() << "size: " << out1.Size()).str());
+    WINML_LOG_COMMENT((std::ostringstream() << "size: " << out1.Size()).str());
     // single key,value pair for each map
     auto iter1 = out1.First();
     auto iter2 = out2.First();
     for (uint32_t i = 0, size = (uint32_t)inputs.size(); i < size; ++i)
     {
-        EXPECT_TRUE(iter2.HasCurrent());
+        WINML_EXPECT_TRUE(iter2.HasCurrent());
         const auto &pair1 = iter1.Current();
         const auto &pair2 = iter2.Current();
-        SCOPED_TRACE((std::ostringstream() << "key: " << pair1.Key().c_str() << ", value " << pair2.Value()).str());
-        EXPECT_TRUE(std::wstring(pair1.Key().c_str()).compare(labels[i]) == 0);
-        EXPECT_TRUE(std::wstring(pair2.Key().c_str()).compare(labels[i]) == 0);
-        EXPECT_TRUE(pair1.Value() == inputs[i] && pair2.Value() == inputs[i]);
+        WINML_LOG_COMMENT((std::ostringstream() << "key: " << pair1.Key().c_str() << ", value " << pair2.Value()).str());
+        WINML_EXPECT_TRUE(std::wstring(pair1.Key().c_str()).compare(labels[i]) == 0);
+        WINML_EXPECT_TRUE(std::wstring(pair2.Key().c_str()).compare(labels[i]) == 0);
+        WINML_EXPECT_TRUE(pair1.Value() == inputs[i] && pair2.Value() == inputs[i]);
         iter1.MoveNext();
         iter2.MoveNext();
     }
-    EXPECT_TRUE(!iter1.HasCurrent());
-    EXPECT_TRUE(!iter2.HasCurrent());
+    WINML_EXPECT_TRUE(!iter1.HasCurrent());
+    WINML_EXPECT_TRUE(!iter2.HasCurrent());
 }
 
-TEST_F(LearningModelBindingAPITestGpu, GpuSqueezeNet)
+static void GpuSqueezeNet()
 {
     std::string gpuInstance("GPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(
@@ -303,7 +306,7 @@ TEST_F(LearningModelBindingAPITestGpu, GpuSqueezeNet)
         /*dataTolerance*/ 0.00001f);
 }
 
-TEST_F(LearningModelBindingAPITestGpu, GpuSqueezeNetEmptyOutputs)
+static void GpuSqueezeNetEmptyOutputs()
 {
     std::string gpuInstance("GPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(
@@ -314,7 +317,7 @@ TEST_F(LearningModelBindingAPITestGpu, GpuSqueezeNetEmptyOutputs)
         OutputBindingStrategy::Empty);
 }
 
-TEST_F(LearningModelBindingAPITestGpu, GpuSqueezeNetUnboundOutputs)
+static void GpuSqueezeNetUnboundOutputs()
 {
     std::string gpuInstance("GPU");
     WinML::Engine::Test::ModelValidator::SqueezeNet(
@@ -326,44 +329,48 @@ TEST_F(LearningModelBindingAPITestGpu, GpuSqueezeNetUnboundOutputs)
 }
 
 // Validates that when the input image is the same as the model expects, the binding step is executed correctly.
-TEST_F(LearningModelBindingAPITestGpu, ImageBindingDimensions)
+static void ImageBindingDimensions()
 {
-    LearningModelBinding m_binding = nullptr;
+    LearningModelBinding learningModelBinding = nullptr;
+    LearningModel learningModel = nullptr;
+    LearningModelSession learningModelSession = nullptr;
+    LearningModelDevice leraningModelDevice = nullptr;
     std::wstring filePath = FileHelpers::GetModulePath() + L"model.onnx";
     // load a model with expected input size: 224 x 224
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::Default));
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromFilePath(filePath));
-    EXPECT_TRUE(m_model != nullptr);
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
-    EXPECT_NO_THROW(m_binding = LearningModelBinding(m_session));
+    WINML_EXPECT_NO_THROW(leraningModelDevice = LearningModelDevice(LearningModelDeviceKind::Default));
+    WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromFilePath(filePath));
+    WINML_EXPECT_TRUE(learningModel != nullptr);
+    WINML_EXPECT_NO_THROW(learningModelSession = LearningModelSession(learningModel, leraningModelDevice));
+    WINML_EXPECT_NO_THROW(learningModelBinding = LearningModelBinding(learningModelSession));
 
     // Create input images and execute bind
     // Test Case 1: both width and height are larger than model expects
     VideoFrame inputImage1(BitmapPixelFormat::Rgba8, 1000, 1000);
     ImageFeatureValue inputTensor = ImageFeatureValue::CreateFromVideoFrame(inputImage1);
-    EXPECT_NO_THROW(m_binding.Bind(L"data_0", inputTensor));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"data_0", inputTensor));
 
     // Test Case 2: only height is larger, while width is smaller
     VideoFrame inputImage2(BitmapPixelFormat::Rgba8, 20, 1000);
     inputTensor = ImageFeatureValue::CreateFromVideoFrame(inputImage2);
-    EXPECT_NO_THROW(m_binding.Bind(L"data_0", inputTensor));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"data_0", inputTensor));
 
     // Test Case 3: only width is larger, while height is smaller
     VideoFrame inputImage3(BitmapPixelFormat::Rgba8, 1000, 20);
     inputTensor = ImageFeatureValue::CreateFromVideoFrame(inputImage3);
-    EXPECT_NO_THROW(m_binding.Bind(L"data_0", inputTensor));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"data_0", inputTensor));
 
     // Test Case 4: both width and height are smaller than model expects
     VideoFrame inputImage4(BitmapPixelFormat::Rgba8, 20, 20);
     inputTensor = ImageFeatureValue::CreateFromVideoFrame(inputImage4);
-    EXPECT_NO_THROW(m_binding.Bind(L"data_0", inputTensor));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"data_0", inputTensor));
 }
 
-TEST_F(LearningModelBindingAPITestGpu, VerifyInvalidBindExceptions)
+static void VerifyInvalidBindExceptions()
 {
-    EXPECT_NO_THROW(LoadModel(L"zipmap-int64.onnx"));
+    LearningModel learningModel = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"zipmap-int64.onnx", learningModel));
 
-    LearningModelSession session(m_model);
+    LearningModelSession session(learningModel);
     LearningModelBinding binding(session);
 
     std::vector<float> inputs = { 0.5f, 0.25f, 0.125f };
@@ -384,47 +391,47 @@ TEST_F(LearningModelBindingAPITestGpu, VerifyInvalidBindExceptions)
 
     // Bind invalid image as tensorfloat input
     auto image = FileHelpers::LoadImageFeatureValue(L"227x227.png");
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"X", image), winrt::hresult_error, ensureWinmlSizeMismatch);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"X", image), winrt::hresult_error, ensureWinmlSizeMismatch);
 
     // Bind invalid map as tensorfloat input
     std::unordered_map<float, float> map;
     auto abiMap = winrt::single_threaded_map(std::move(map));
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"X", abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"X", abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid sequence as tensorfloat input
     std::vector<uint32_t> sequence;
     auto abiSequence = winrt::single_threaded_vector(std::move(sequence));
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"X", abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"X", abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid tensor size as tensorfloat input
     auto tensorBoolean = TensorBoolean::Create();
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"X", tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"X", tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid tensor shape as tensorfloat input
     auto tensorInvalidShape = TensorFloat::Create(std::vector<int64_t> { 2, 3, 4 });
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"X", tensorInvalidShape), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"X", tensorInvalidShape), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     /*
         Verify sequence bindings throw correct bind exceptions
     */
 
     // Bind invalid image as sequence<map<int, float> output
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", image), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", image), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid map as sequence<map<int, float> output
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid sequence<int> as sequence<map<int, float> output
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid tensor as sequence<map<int, float> output
-    EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(binding.Bind(L"Y", tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     /*
         Verify image bindings throw correct bind exceptions
     */
 
-    // EXPECT_NO_THROW(LoadModel(L"fns-candy.onnx"));
+    // WINML_EXPECT_NO_THROW(LoadModel(L"fns-candy.onnx"));
 
     // LearningModelSession imageSession(m_model);
     // LearningModelBinding imageBinding(imageSession);
@@ -432,74 +439,77 @@ TEST_F(LearningModelBindingAPITestGpu, VerifyInvalidBindExceptions)
     // auto inputName = m_model.InputFeatures().First().Current().Name();
 
     // // Bind invalid map as image input
-    // EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
+    // WINML_EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // // Bind invalid sequence as image input
-    // EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
+    // WINML_EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // // Bind invalid tensor type as image input
-    // EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
+    // WINML_EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // // Bind invalid tensor size as image input
     // auto tensorFloat = TensorFloat::Create(std::vector<int64_t> { 1, 1, 100, 100 });
-    // EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, tensorFloat), winrt::hresult_error, ensureWinmlInvalidBinding);
+    // WINML_EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, tensorFloat), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // // Bind invalid tensor shape as image input
-    // EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, tensorInvalidShape), winrt::hresult_error, ensureWinmlInvalidBinding);
+    // WINML_EXPECT_THROW_SPECIFIC(imageBinding.Bind(inputName, tensorInvalidShape), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     /*
         Verify map bindings throw correct bind exceptions
     */
-    EXPECT_NO_THROW(LoadModel(L"dictvectorizer-int64.onnx"));
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"dictvectorizer-int64.onnx", learningModel));
 
-    LearningModelSession mapSession(m_model);
+    LearningModelSession mapSession(learningModel);
     LearningModelBinding mapBinding(mapSession);
 
-    auto inputName = m_model.InputFeatures().First().Current().Name();
+    auto inputName = learningModel.InputFeatures().First().Current().Name();
 
     // Bind invalid image as image input
     auto smallImage = FileHelpers::LoadImageFeatureValue(L"100x100.png");
-    EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, smallImage), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, smallImage), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid map as image input
-    EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, abiMap), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid sequence as image input
-    EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, abiSequence), winrt::hresult_error, ensureWinmlInvalidBinding);
 
     // Bind invalid tensor type as image input
-    EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
+    WINML_EXPECT_THROW_SPECIFIC(mapBinding.Bind(inputName, tensorBoolean), winrt::hresult_error, ensureWinmlInvalidBinding);
 }
 
 // Verify that it throws an error when binding an invalid name.
-TEST_F(LearningModelBindingAPITestGpu, BindInvalidInputName)
+static void BindInvalidInputName()
 {
-    LearningModelBinding m_binding = nullptr;
+    LearningModel learningModel = nullptr;
+    LearningModelBinding learningModelBinding = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    LearningModelSession learningModelSession = nullptr;
     std::wstring modelPath = FileHelpers::GetModulePath() + L"Add_ImageNet1920.onnx";
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromFilePath(modelPath));
-    EXPECT_TRUE(m_model != nullptr);
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::Default));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
-    EXPECT_NO_THROW(m_binding = LearningModelBinding(m_session));
+    WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromFilePath(modelPath));
+    WINML_EXPECT_TRUE(learningModel != nullptr);
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::Default));
+    WINML_EXPECT_NO_THROW(learningModelSession = LearningModelSession(learningModel, learningModelDevice));
+    WINML_EXPECT_NO_THROW(learningModelBinding = LearningModelBinding(learningModelSession));
 
     VideoFrame iuputImage(BitmapPixelFormat::Rgba8, 1920, 1080);
     ImageFeatureValue inputTensor = ImageFeatureValue::CreateFromVideoFrame(iuputImage);
 
-    auto first = m_model.InputFeatures().First();
+    auto first = learningModel.InputFeatures().First();
     std::wstring testInvalidName = L"0";
 
     // Verify that testInvalidName is not in model's InputFeatures
     while (first.HasCurrent())
     {
-        EXPECT_NE(testInvalidName, first.Current().Name());
+        WINML_EXPECT_NOT_EQUAL(testInvalidName, first.Current().Name());
         first.MoveNext();
     }
 
     // Bind inputTensor to a valid input name
-    EXPECT_NO_THROW(m_binding.Bind(L"input_39:0", inputTensor));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"input_39:0", inputTensor));
 
     // Bind inputTensor to an invalid input name
-    EXPECT_THROW_SPECIFIC(m_binding.Bind(testInvalidName, inputTensor),
+    WINML_EXPECT_THROW_SPECIFIC(learningModelBinding.Bind(testInvalidName, inputTensor),
         winrt::hresult_error,
         [](const winrt::hresult_error& e) -> bool
     {
@@ -507,15 +517,18 @@ TEST_F(LearningModelBindingAPITestGpu, BindInvalidInputName)
     });
 }
 
-TEST_F(LearningModelBindingAPITest, VerifyOutputAfterEvaluateAsyncCalledTwice)
+static void VerifyOutputAfterEvaluateAsyncCalledTwice()
 {
-    LearningModelBinding m_binding = nullptr;
+    LearningModel learningModel = nullptr;
+    LearningModelBinding learningModelBinding = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    LearningModelSession learningModelSession = nullptr;
     std::wstring filePath = FileHelpers::GetModulePath() + L"relu.onnx";
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::Default));
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromFilePath(filePath));
-    EXPECT_TRUE(m_model != nullptr);
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
-    EXPECT_NO_THROW(m_binding = LearningModelBinding(m_session));
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::Default));
+    WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromFilePath(filePath));
+    WINML_EXPECT_TRUE(learningModel != nullptr);
+    WINML_EXPECT_NO_THROW(learningModelSession = LearningModelSession(learningModel, learningModelDevice));
+    WINML_EXPECT_NO_THROW(learningModelBinding = LearningModelBinding(learningModelSession));
 
     auto inputShape = std::vector<int64_t>{ 5 };
     auto inputData1 = std::vector<float>{ -50.f, -25.f, 0.f, 25.f, 50.f };
@@ -530,22 +543,22 @@ TEST_F(LearningModelBindingAPITest, VerifyOutputAfterEvaluateAsyncCalledTwice)
             inputShape,
             single_threaded_vector<float>(std::move(inputData2)).GetView());
 
-    EXPECT_NO_THROW(m_binding.Bind(L"X", inputValue1));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"X", inputValue1));
 
     auto outputValue = TensorFloat::Create();
-    EXPECT_NO_THROW(m_binding.Bind(L"Y", outputValue));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"Y", outputValue));
 
-    EXPECT_NO_THROW(m_session.Evaluate(m_binding, L""));
+    WINML_EXPECT_NO_THROW(learningModelSession.Evaluate(learningModelBinding, L""));
 
     auto buffer1 = outputValue.GetAsVectorView();
-    EXPECT_TRUE(buffer1 != nullptr);
+    WINML_EXPECT_TRUE(buffer1 != nullptr);
 
     // The second evaluation
     // If we don't bind output again, the output value will not change
-    EXPECT_NO_THROW(m_binding.Bind(L"X", inputValue2));
-    EXPECT_NO_THROW(m_session.Evaluate(m_binding, L""));
+    WINML_EXPECT_NO_THROW(learningModelBinding.Bind(L"X", inputValue2));
+    WINML_EXPECT_NO_THROW(learningModelSession.Evaluate(learningModelBinding, L""));
     auto buffer2 = outputValue.GetAsVectorView();
-    EXPECT_EQ(buffer1.Size(), buffer2.Size());
+    WINML_EXPECT_EQUAL(buffer1.Size(), buffer2.Size());
     bool isSame = true;
     for (uint32_t i = 0; i < buffer1.Size(); ++i)
     {
@@ -555,7 +568,7 @@ TEST_F(LearningModelBindingAPITest, VerifyOutputAfterEvaluateAsyncCalledTwice)
             break;
         }
     }
-    EXPECT_FALSE(isSame);
+    WINML_EXPECT_FALSE(isSame);
 }
 
 static VideoFrame CreateVideoFrame(const wchar_t* path)
@@ -567,7 +580,7 @@ static VideoFrame CreateVideoFrame(const wchar_t* path)
     return VideoFrame::CreateWithSoftwareBitmap(softwareBitmap);
 }
 
-TEST_F(LearningModelBindingAPITest, VerifyOutputAfterImageBindCalledTwice)
+static void VerifyOutputAfterImageBindCalledTwice()
 {
     std::wstring fullModelPath = FileHelpers::GetModulePath() + L"model.onnx";
     std::wstring fullImagePath1 = FileHelpers::GetModulePath() + L"kitten_224.png";
@@ -575,9 +588,9 @@ TEST_F(LearningModelBindingAPITest, VerifyOutputAfterImageBindCalledTwice)
 
     // winml model creation
     LearningModel model = nullptr;
-    EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
+    WINML_EXPECT_NO_THROW(model = LearningModel::LoadFromFilePath(fullModelPath));
     LearningModelSession modelSession = nullptr;
-    EXPECT_NO_THROW(modelSession = LearningModelSession(model, LearningModelDevice(LearningModelDeviceKind::Default)));
+    WINML_EXPECT_NO_THROW(modelSession = LearningModelSession(model, LearningModelDevice(LearningModelDeviceKind::Default)));
     LearningModelBinding modelBinding(modelSession);
 
     // create the tensor for the actual output
@@ -587,8 +600,8 @@ TEST_F(LearningModelBindingAPITest, VerifyOutputAfterImageBindCalledTwice)
     // Bind image 1 and evaluate
     auto frame = CreateVideoFrame(fullImagePath1.c_str());
     auto imageTensor = ImageFeatureValue::CreateFromVideoFrame(frame);
-    EXPECT_NO_THROW(modelBinding.Bind(L"data_0", imageTensor));
-    EXPECT_NO_THROW(modelSession.Evaluate(modelBinding, L""));
+    WINML_EXPECT_NO_THROW(modelBinding.Bind(L"data_0", imageTensor));
+    WINML_EXPECT_NO_THROW(modelSession.Evaluate(modelBinding, L""));
 
     // Store 1st result
     auto outputVectorView1 = output.GetAsVectorView();
@@ -598,13 +611,13 @@ TEST_F(LearningModelBindingAPITest, VerifyOutputAfterImageBindCalledTwice)
     // The expected result is that the videoframe will be re-tensorized at bind
     auto frame2 = CreateVideoFrame(fullImagePath2.c_str());
     frame2.CopyToAsync(frame).get();
-    EXPECT_NO_THROW(modelBinding.Bind(L"data_0", imageTensor));
-    EXPECT_NO_THROW(modelSession.Evaluate(modelBinding, L""));
+    WINML_EXPECT_NO_THROW(modelBinding.Bind(L"data_0", imageTensor));
+    WINML_EXPECT_NO_THROW(modelSession.Evaluate(modelBinding, L""));
 
     // Store 2nd result
     auto outputVectorView2 = output.GetAsVectorView();
 
-    EXPECT_EQ(outputVectorView1.Size(), outputVectorView2.Size());
+    WINML_EXPECT_EQUAL(outputVectorView1.Size(), outputVectorView2.Size());
     bool isSame = true;
     for (uint32_t i = 0; i < outputVectorView1.Size(); ++i)
     {
@@ -614,5 +627,32 @@ TEST_F(LearningModelBindingAPITest, VerifyOutputAfterImageBindCalledTwice)
             break;
         }
     }
-    EXPECT_FALSE(isSame);
+    WINML_EXPECT_FALSE(isSame);
 }
+
+const LearningModelBindingAPITestApi& getapi() {
+  static constexpr LearningModelBindingAPITestApi api =
+  {
+    LearningModelBindingAPITestSetup,
+    LearningModelBindingAPITestGpuSetup,
+    CpuSqueezeNet,
+    CpuSqueezeNetEmptyOutputs,
+    CpuSqueezeNetUnboundOutputs,
+    CpuSqueezeNetBindInputTensorAsInspectable,
+    CastMapInt64,
+    DictionaryVectorizerMapInt64,
+    DictionaryVectorizerMapString,
+    ZipMapInt64,
+    ZipMapInt64Unbound,
+    ZipMapString,
+    GpuSqueezeNet,
+    GpuSqueezeNetEmptyOutputs,
+    GpuSqueezeNetUnboundOutputs,
+    ImageBindingDimensions,
+    VerifyInvalidBindExceptions,
+    BindInvalidInputName,
+    VerifyOutputAfterEvaluateAsyncCalledTwice,
+    VerifyOutputAfterImageBindCalledTwice
+  };
+  return api;
+}
\ No newline at end of file
diff --git a/winml/test/api/LearningModelBindingAPITest.h b/winml/test/api/LearningModelBindingAPITest.h
new file mode 100644
index 0000000000000..c93be87c3cb60
--- /dev/null
+++ b/winml/test/api/LearningModelBindingAPITest.h
@@ -0,0 +1,49 @@
+#include "test.h"
+
+struct LearningModelBindingAPITestApi {
+  SetupTest LearningModelBindingAPITestSetup;
+  SetupTest LearningModelBindingAPITestGpuSetup;
+  VoidTest CpuSqueezeNet;
+  VoidTest CpuSqueezeNetEmptyOutputs;
+  VoidTest CpuSqueezeNetUnboundOutputs;
+  VoidTest CpuSqueezeNetBindInputTensorAsInspectable;
+  VoidTest CastMapInt64;
+  VoidTest DictionaryVectorizerMapInt64;
+  VoidTest DictionaryVectorizerMapString;
+  VoidTest ZipMapInt64;
+  VoidTest ZipMapInt64Unbound;
+  VoidTest ZipMapString;
+  VoidTest GpuSqueezeNet;
+  VoidTest GpuSqueezeNetEmptyOutputs;
+  VoidTest GpuSqueezeNetUnboundOutputs;
+  VoidTest ImageBindingDimensions;
+  VoidTest VerifyInvalidBindExceptions;
+  VoidTest BindInvalidInputName;
+  VoidTest VerifyOutputAfterEvaluateAsyncCalledTwice;
+  VoidTest VerifyOutputAfterImageBindCalledTwice;
+};
+const LearningModelBindingAPITestApi& getapi();
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelBindingAPITest, LearningModelBindingAPITestSetup)
+WINML_TEST(LearningModelBindingAPITest, CpuSqueezeNet)
+WINML_TEST(LearningModelBindingAPITest, CpuSqueezeNetEmptyOutputs)
+WINML_TEST(LearningModelBindingAPITest, CpuSqueezeNetUnboundOutputs)
+WINML_TEST(LearningModelBindingAPITest, CpuSqueezeNetBindInputTensorAsInspectable)
+WINML_TEST(LearningModelBindingAPITest, CastMapInt64)
+WINML_TEST(LearningModelBindingAPITest, DictionaryVectorizerMapInt64)
+WINML_TEST(LearningModelBindingAPITest, DictionaryVectorizerMapString)
+WINML_TEST(LearningModelBindingAPITest, ZipMapInt64)
+WINML_TEST(LearningModelBindingAPITest, ZipMapInt64Unbound)
+WINML_TEST(LearningModelBindingAPITest, ZipMapString)
+WINML_TEST(LearningModelBindingAPITest, VerifyOutputAfterEvaluateAsyncCalledTwice)
+WINML_TEST(LearningModelBindingAPITest, VerifyOutputAfterImageBindCalledTwice)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelBindingAPITestGpu, LearningModelBindingAPITestGpuSetup)
+WINML_TEST(LearningModelBindingAPITestGpu, GpuSqueezeNet)
+WINML_TEST(LearningModelBindingAPITestGpu, GpuSqueezeNetEmptyOutputs)
+WINML_TEST(LearningModelBindingAPITestGpu, GpuSqueezeNetUnboundOutputs)
+WINML_TEST(LearningModelBindingAPITestGpu, ImageBindingDimensions)
+WINML_TEST(LearningModelBindingAPITestGpu, VerifyInvalidBindExceptions)
+WINML_TEST(LearningModelBindingAPITestGpu, BindInvalidInputName)
+WINML_TEST_CLASS_END()
\ No newline at end of file
diff --git a/winml/test/api/LearningModelSessionAPITest.cpp b/winml/test/api/LearningModelSessionAPITest.cpp
index bc6494a3a71ac..6ee0b3d57f751 100644
--- a/winml/test/api/LearningModelSessionAPITest.cpp
+++ b/winml/test/api/LearningModelSessionAPITest.cpp
@@ -1,6 +1,6 @@
 #include "testPch.h"
 #include "APITest.h"
-
+#include "LearningModelSessionAPITest.h"
 #include "winrt/Windows.Storage.h"
 
 #include "DeviceHelpers.h"
@@ -16,144 +16,151 @@ using namespace winrt::Windows::Foundation::Collections;
 
 using winrt::Windows::Foundation::IPropertyValue;
 
-class LearningModelSessionAPITests : public APITest
-{};
+static void LearningModelSessionAPITestSetup() {
+  init_apartment();
+}
 
-class LearningModelSessionAPITestsGpu : public APITest
-{
-protected:
-    void SetUp() override
-    {
-        GPUTEST
-    }
-};
+static void LearningModelSessionAPITestGpuSetup() {
+  GPUTEST;
+  init_apartment();
+}
 
-class LearningModelSessionAPITestsSkipEdgeCore : public LearningModelSessionAPITestsGpu
-{
-protected:
-    void SetUp() override
-    {
-        LearningModelSessionAPITestsGpu::SetUp();
-        SKIP_EDGECORE
-    }
-};
+static void LearningModelSessionAPITestsSkipEdgeCoreSetup() {
+  LearningModelSessionAPITestGpuSetup();
+  SKIP_EDGECORE
+}
 
-TEST_F(LearningModelSessionAPITests, CreateSessionDeviceDefault)
+static void CreateSessionDeviceDefault()
 {
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+    LearningModel learningModel = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
 
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::Default));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::Default));
+    WINML_EXPECT_NO_THROW(LearningModelSession(learningModel, learningModelDevice));
 }
 
-TEST_F(LearningModelSessionAPITests, CreateSessionDeviceCpu)
+static void CreateSessionDeviceCpu()
 {
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+    LearningModel learningModel = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
 
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::Cpu));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::Cpu));
+    WINML_EXPECT_NO_THROW(LearningModelSession(learningModel, learningModelDevice));
     // for the CPU device, make sure that we get back NULL and 0 for any device properties
-    EXPECT_FALSE(m_device.Direct3D11Device());
+    WINML_EXPECT_EQUAL(learningModelDevice.Direct3D11Device(), nullptr);
     LARGE_INTEGER id;
-    id.QuadPart = GetAdapterIdQuadPart();
-    EXPECT_EQ(id.LowPart, static_cast<DWORD>(0));
-    EXPECT_EQ(id.HighPart, 0);
+    id.QuadPart = APITest::GetAdapterIdQuadPart(learningModelDevice);
+    WINML_EXPECT_EQUAL(id.LowPart, static_cast<DWORD>(0));
+    WINML_EXPECT_EQUAL(id.HighPart, 0);
 }
 
-TEST_F(LearningModelSessionAPITests, CreateSessionWithModelLoadedFromStream)
+static void CreateSessionWithModelLoadedFromStream()
 {
+    LearningModel learningModel = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
     std::wstring path = FileHelpers::GetModulePath() + L"model.onnx";
     auto storageFile = winrt::Windows::Storage::StorageFile::GetFileFromPathAsync(path).get();
 
-    EXPECT_NO_THROW(m_model = LearningModel::LoadFromStream(storageFile));
+    WINML_EXPECT_NO_THROW(learningModel = LearningModel::LoadFromStream(storageFile));
 
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::Default));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::Default));
+    WINML_EXPECT_NO_THROW(LearningModelSession(learningModel, learningModelDevice));
 }
 
-TEST_F(LearningModelSessionAPITestsGpu, CreateSessionDeviceDirectX)
+static void CreateSessionDeviceDirectX()
 {
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+    LearningModel learningModel = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
 
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::DirectX));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::DirectX));
+    WINML_EXPECT_NO_THROW(LearningModelSession(learningModel, learningModelDevice));
 }
 
-TEST_F(LearningModelSessionAPITestsGpu, CreateSessionDeviceDirectXHighPerformance)
+static void CreateSessionDeviceDirectXHighPerformance()
 {
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+    LearningModel learningModel = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
 
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::DirectXHighPerformance));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
+    WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::DirectXHighPerformance));
+    WINML_EXPECT_NO_THROW(LearningModelSession(learningModel, learningModelDevice));
 }
 
-TEST_F(LearningModelSessionAPITestsGpu, CreateSessionDeviceDirectXMinimumPower)
+static void CreateSessionDeviceDirectXMinimumPower()
 {
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+  LearningModel learningModel = nullptr;
+  LearningModelDevice learningModelDevice = nullptr;
+  WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
 
-    EXPECT_NO_THROW(m_device = LearningModelDevice(LearningModelDeviceKind::DirectXMinPower));
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
+  WINML_EXPECT_NO_THROW(learningModelDevice = LearningModelDevice(LearningModelDeviceKind::DirectXMinPower));
+  WINML_EXPECT_NO_THROW(LearningModelSession(learningModel, learningModelDevice));
 }
 
-TEST_F(LearningModelSessionAPITestsSkipEdgeCore, AdapterIdAndDevice)
-{
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+static void AdapterIdAndDevice() {
+    LearningModel learningModel = nullptr;
+    LearningModelDevice learningModelDevice = nullptr;
+    LearningModelSession learningModelSession = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
 
     com_ptr<IDXGIFactory6> factory;
-    EXPECT_HRESULT_SUCCEEDED(CreateDXGIFactory1(__uuidof(IDXGIFactory6), factory.put_void()));
+    WINML_EXPECT_HRESULT_SUCCEEDED(CreateDXGIFactory1(__uuidof(IDXGIFactory6), factory.put_void()));
     com_ptr<IDXGIAdapter> adapter;
 
-    m_device = LearningModelDevice(LearningModelDeviceKind::DirectX);
-    EXPECT_HRESULT_SUCCEEDED(factory->EnumAdapters(0, adapter.put()));
+    learningModelDevice = LearningModelDevice(LearningModelDeviceKind::DirectX);
+    WINML_EXPECT_HRESULT_SUCCEEDED(factory->EnumAdapters(0, adapter.put()));
     DXGI_ADAPTER_DESC desc;
-    EXPECT_HRESULT_SUCCEEDED(adapter->GetDesc(&desc));
+    WINML_EXPECT_HRESULT_SUCCEEDED(adapter->GetDesc(&desc));
     LARGE_INTEGER id;
-    id.QuadPart = GetAdapterIdQuadPart();
-    EXPECT_EQ(desc.AdapterLuid.LowPart, id.LowPart);
-    EXPECT_EQ(desc.AdapterLuid.HighPart, id.HighPart);
-    EXPECT_TRUE(m_device.Direct3D11Device() != nullptr);
+    id.QuadPart = APITest::GetAdapterIdQuadPart(learningModelDevice);
+    WINML_EXPECT_EQUAL(desc.AdapterLuid.LowPart, id.LowPart);
+    WINML_EXPECT_EQUAL(desc.AdapterLuid.HighPart, id.HighPart);
+    WINML_EXPECT_TRUE(learningModelDevice.Direct3D11Device() != nullptr);
 
-    m_device = LearningModelDevice(LearningModelDeviceKind::DirectXHighPerformance);
+    learningModelDevice = LearningModelDevice(LearningModelDeviceKind::DirectXHighPerformance);
     adapter = nullptr;
-    EXPECT_HRESULT_SUCCEEDED(factory->EnumAdapterByGpuPreference(0, DXGI_GPU_PREFERENCE_HIGH_PERFORMANCE, __uuidof(IDXGIAdapter), adapter.put_void()));
-    EXPECT_HRESULT_SUCCEEDED(adapter->GetDesc(&desc));
-    id.QuadPart = GetAdapterIdQuadPart();
-    EXPECT_EQ(desc.AdapterLuid.LowPart, id.LowPart);
-    EXPECT_EQ(desc.AdapterLuid.HighPart, id.HighPart);
-    EXPECT_TRUE(m_device.Direct3D11Device() != nullptr);
+    WINML_EXPECT_HRESULT_SUCCEEDED(factory->EnumAdapterByGpuPreference(0, DXGI_GPU_PREFERENCE_HIGH_PERFORMANCE, __uuidof(IDXGIAdapter), adapter.put_void()));
+    WINML_EXPECT_HRESULT_SUCCEEDED(adapter->GetDesc(&desc));
+    id.QuadPart = APITest::GetAdapterIdQuadPart(learningModelDevice);
+    WINML_EXPECT_EQUAL(desc.AdapterLuid.LowPart, id.LowPart);
+    WINML_EXPECT_EQUAL(desc.AdapterLuid.HighPart, id.HighPart);
+    WINML_EXPECT_TRUE(learningModelDevice.Direct3D11Device() != nullptr);
 
     adapter = nullptr;
-    m_device = LearningModelDevice(LearningModelDeviceKind::DirectXMinPower);
-    EXPECT_HRESULT_SUCCEEDED(factory->EnumAdapterByGpuPreference(0, DXGI_GPU_PREFERENCE_MINIMUM_POWER, __uuidof(IDXGIAdapter), adapter.put_void()));
-    EXPECT_HRESULT_SUCCEEDED(adapter->GetDesc(&desc));
-    id.QuadPart = GetAdapterIdQuadPart();
-    EXPECT_EQ(desc.AdapterLuid.LowPart, id.LowPart);
-    EXPECT_EQ(desc.AdapterLuid.HighPart, id.HighPart);
-    EXPECT_TRUE(m_device.Direct3D11Device() != nullptr);
-
-    EXPECT_NO_THROW(m_session = LearningModelSession(m_model, m_device));
-    EXPECT_EQ(m_session.Device().AdapterId(), m_device.AdapterId());
+    learningModelDevice = LearningModelDevice(LearningModelDeviceKind::DirectXMinPower);
+    WINML_EXPECT_HRESULT_SUCCEEDED(factory->EnumAdapterByGpuPreference(0, DXGI_GPU_PREFERENCE_MINIMUM_POWER, __uuidof(IDXGIAdapter), adapter.put_void()));
+    WINML_EXPECT_HRESULT_SUCCEEDED(adapter->GetDesc(&desc));
+    id.QuadPart = APITest::GetAdapterIdQuadPart(learningModelDevice);
+    WINML_EXPECT_EQUAL(desc.AdapterLuid.LowPart, id.LowPart);
+    WINML_EXPECT_EQUAL(desc.AdapterLuid.HighPart, id.HighPart);
+    WINML_EXPECT_TRUE(learningModelDevice.Direct3D11Device() != nullptr);
+
+    WINML_EXPECT_NO_THROW(learningModelSession = LearningModelSession(learningModel, learningModelDevice));
+    WINML_EXPECT_EQUAL(learningModelSession.Device().AdapterId(), learningModelDevice.AdapterId());
 }
 
-TEST_F(LearningModelSessionAPITests, EvaluateFeatures)
+static void EvaluateFeatures()
 {
     std::vector<int64_t> shape = { 4 };
     std::vector<winrt::hstring> data = { L"one", L"two", L"three", L"four" };
 
     // create from buffer
     auto tensor = TensorString::CreateFromArray(shape, data);
-    EXPECT_EQ(tensor.GetAsVectorView().Size(), data.size());
-    EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
+    WINML_EXPECT_EQUAL(tensor.GetAsVectorView().Size(), data.size());
+    WINML_EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
 
     // create from vector view
     auto dataCopy = data;
     tensor = TensorString::CreateFromIterable(
         shape, winrt::single_threaded_vector<winrt::hstring>(std::move(dataCopy)).GetView());
-    EXPECT_EQ(tensor.GetAsVectorView().Size(), data.size());
-    EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
+    WINML_EXPECT_EQUAL(tensor.GetAsVectorView().Size(), data.size());
+    WINML_EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
 
-    EXPECT_NO_THROW(LoadModel(L"id-tensor-string.onnx"));
-    LearningModelSession session(m_model);
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"id-tensor-string.onnx", learningModel));
+    LearningModelSession session(learningModel);
 
     auto outputTensor = TensorString::Create();
 
@@ -164,29 +171,30 @@ TEST_F(LearningModelSessionAPITests, EvaluateFeatures)
     session.EvaluateFeatures(featureswinrtmap, L"0");
 
     // verify identity model round-trip works
-    EXPECT_EQ(outputTensor.GetAsVectorView().Size(), data.size());
-    EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(outputTensor.GetAsVectorView())));
+    WINML_EXPECT_EQUAL(outputTensor.GetAsVectorView().Size(), data.size());
+    WINML_EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(outputTensor.GetAsVectorView())));
 }
 
-TEST_F(LearningModelSessionAPITests, EvaluateFeaturesAsync)
+static void EvaluateFeaturesAsync()
 {
     std::vector<int64_t> shape = { 4 };
     std::vector<winrt::hstring> data = { L"one", L"two", L"three", L"four" };
 
     // create from buffer
     auto tensor = TensorString::CreateFromArray(shape, data);
-    EXPECT_EQ(tensor.GetAsVectorView().Size(), data.size());
-    EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
+    WINML_EXPECT_EQUAL(tensor.GetAsVectorView().Size(), data.size());
+    WINML_EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
 
     // create from vector view
     auto dataCopy = data;
     tensor = TensorString::CreateFromIterable(
         shape, winrt::single_threaded_vector<winrt::hstring>(std::move(dataCopy)).GetView());
-    EXPECT_EQ(tensor.GetAsVectorView().Size(), data.size());
-    EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
+    WINML_EXPECT_EQUAL(tensor.GetAsVectorView().Size(), data.size());
+    WINML_EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(tensor.GetAsVectorView())));
 
-    EXPECT_NO_THROW(LoadModel(L"id-tensor-string.onnx"));
-    LearningModelSession session(m_model);
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"id-tensor-string.onnx", learningModel));
+    LearningModelSession session(learningModel);
 
     auto outputTensor = TensorString::Create(shape);
 
@@ -197,37 +205,39 @@ TEST_F(LearningModelSessionAPITests, EvaluateFeaturesAsync)
     session.EvaluateFeaturesAsync(featureswinrtmap, L"0").get();
 
     // verify identity model round-trip works
-    EXPECT_EQ(outputTensor.GetAsVectorView().Size(), data.size());
-    EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(outputTensor.GetAsVectorView())));
+    WINML_EXPECT_EQUAL(outputTensor.GetAsVectorView().Size(), data.size());
+    WINML_EXPECT_TRUE(std::equal(data.cbegin(), data.cend(), begin(outputTensor.GetAsVectorView())));
 }
 
-TEST_F(LearningModelSessionAPITests, EvaluationProperties)
+static void EvaluationProperties()
 {
     // load a model
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
     // create a session
-    m_session = LearningModelSession(m_model);
+    LearningModelSession learningModelSession = nullptr;
+    learningModelSession = LearningModelSession(learningModel);
     // set a property
     auto value = winrt::Windows::Foundation::PropertyValue::CreateBoolean(true);
-    m_session.EvaluationProperties().Insert(L"propName1", value);
+    learningModelSession.EvaluationProperties().Insert(L"propName1", value);
     // get the property and make sure it's there with the right value
-    auto value2 = m_session.EvaluationProperties().Lookup(L"propName1");
-    EXPECT_EQ(value2.as<IPropertyValue>().GetBoolean(), true);
+    auto value2 = learningModelSession.EvaluationProperties().Lookup(L"propName1");
+    WINML_EXPECT_EQUAL(value2.as<IPropertyValue>().GetBoolean(), true);
 }
 
 static LearningModelSession CreateSession(LearningModel model)
 {
     LearningModelDevice device(nullptr);
-    EXPECT_NO_THROW(device = LearningModelDevice(LearningModelDeviceKind::DirectX));
+    WINML_EXPECT_NO_THROW(device = LearningModelDevice(LearningModelDeviceKind::DirectX));
 
     LearningModelSession session(nullptr);
     if (DeviceHelpers::IsFloat16Supported(device))
     {
-        EXPECT_NO_THROW(session = LearningModelSession(model, device));
+        WINML_EXPECT_NO_THROW(session = LearningModelSession(model, device));
     }
     else
     {
-        EXPECT_THROW_SPECIFIC(
+        WINML_EXPECT_THROW_SPECIFIC(
             session = LearningModelSession(model, device),
             winrt::hresult_error,
             [](const winrt::hresult_error& e) -> bool
@@ -239,26 +249,28 @@ static LearningModelSession CreateSession(LearningModel model)
     return session;
 }
 
-TEST_F(LearningModelSessionAPITestsGpu, CreateSessionWithCastToFloat16InModel)
+static void CreateSessionWithCastToFloat16InModel()
 {
     // load a model
-    EXPECT_NO_THROW(LoadModel(L"fp16-truncate-with-cast.onnx"));
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"fp16-truncate-with-cast.onnx", learningModel));
 
-    CreateSession(m_model);
+    CreateSession(learningModel);
 }
 
-TEST_F(LearningModelSessionAPITestsGpu, DISABLED_CreateSessionWithFloat16InitializersInModel)
+static void DISABLED_CreateSessionWithFloat16InitializersInModel()
 {
     // Disabled due to https://microsoft.visualstudio.com/DefaultCollection/OS/_workitems/edit/21624720:
     // Model fails to resolve due to ORT using incorrect IR version within partition
 
     // load a model
-    EXPECT_NO_THROW(LoadModel(L"fp16-initializer.onnx"));
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"fp16-initializer.onnx", learningModel));
 
-    CreateSession(m_model);
+    CreateSession(learningModel);
 }
 
-static void EvaluateSessionAndCloseModel(
+static void EvaluateSessionAndCloseModelHelper(
     LearningModelDeviceKind kind,
     bool close_model_on_session_creation)
 {
@@ -275,7 +287,7 @@ static void EvaluateSessionAndCloseModel(
     // ensure you can create a session from the model
     LearningModelSession session(nullptr);
 
-    EXPECT_NO_THROW(session = LearningModelSession(model, device, options));
+    WINML_EXPECT_NO_THROW(session = LearningModelSession(model, device, options));
 
     std::vector<float> input(1000);
     std::iota(std::begin(input), std::end(input), 0.0f);
@@ -284,12 +296,12 @@ static void EvaluateSessionAndCloseModel(
     binding.Bind(L"input", tensor_input);
 
     LearningModelEvaluationResult result(nullptr);
-    EXPECT_NO_THROW(result = session.Evaluate(binding, L""));
+    WINML_EXPECT_NO_THROW(result = session.Evaluate(binding, L""));
 
     if (close_model_on_session_creation)
     {
         // ensure that the model has been closed
-        EXPECT_THROW_SPECIFIC(
+        WINML_EXPECT_THROW_SPECIFIC(
             LearningModelSession(model, device, options),
             winrt::hresult_error,
             [](const winrt::hresult_error& e) -> bool
@@ -299,19 +311,20 @@ static void EvaluateSessionAndCloseModel(
     }
     else
     {
-        EXPECT_NO_THROW(LearningModelSession(model, device, options));
+        WINML_EXPECT_NO_THROW(LearningModelSession(model, device, options));
     }
 }
 
-TEST_F(LearningModelSessionAPITests, EvaluateSessionAndCloseModel)
+static void EvaluateSessionAndCloseModel()
 {
-    EXPECT_NO_THROW(::EvaluateSessionAndCloseModel(LearningModelDeviceKind::Cpu, true));
-    EXPECT_NO_THROW(::EvaluateSessionAndCloseModel(LearningModelDeviceKind::Cpu, false));
+    WINML_EXPECT_NO_THROW(::EvaluateSessionAndCloseModelHelper(LearningModelDeviceKind::Cpu, true));
+    WINML_EXPECT_NO_THROW(::EvaluateSessionAndCloseModelHelper(LearningModelDeviceKind::Cpu, false));
 }
 
-TEST_F(LearningModelSessionAPITests, CloseSession)
+static void CloseSession()
 {
-    EXPECT_NO_THROW(LoadModel(L"model.onnx"));
+    LearningModel learningModel = nullptr;
+    WINML_EXPECT_NO_THROW(APITest::LoadModel(L"model.onnx", learningModel));
     LearningModelSession session = nullptr;
 
     /*
@@ -329,7 +342,7 @@ TEST_F(LearningModelSessionAPITests, CloseSession)
     SIZE_T afterSessionCloseWorkingSetSize = 0;
     bool getProcessMemoryInfoSuccess = false;
     */
-    EXPECT_NO_THROW(session = LearningModelSession(m_model));
+    WINML_EXPECT_NO_THROW(session = LearningModelSession(learningModel));
 
     /*
     // Get the current process memory info after session creation.
@@ -341,7 +354,7 @@ TEST_F(LearningModelSessionAPITests, CloseSession)
     beforeSessionCloseWorkingSetSize = pmc.WorkingSetSize;
     pmc = { 0 };
     */
-    EXPECT_NO_THROW(session.Close());
+    WINML_EXPECT_NO_THROW(session.Close());
 
     /*
     Bug 23659026: Working set difference tolerance is too tight for LearningModelSessionAPITests::CloseSession
@@ -367,17 +380,41 @@ TEST_F(LearningModelSessionAPITests, CloseSession)
     */
 
     // verify that model still has metadata info after session close
-    std::wstring author(m_model.Author());
-    EXPECT_EQ(author, L"onnx-caffe2");
+    std::wstring author(learningModel.Author());
+    WINML_EXPECT_EQUAL(author, L"onnx-caffe2");
 
     // verify that session throws RO_E_CLOSED error
     std::vector<float> input(1 * 3 * 224 * 224, 0);
     std::vector<int64_t> shape = { 1, 3, 224, 224 };
     auto tensor_input = TensorFloat::CreateFromShapeArrayAndDataArray(shape, input);
-    EXPECT_THROW_SPECIFIC(LearningModelBinding binding(session),
+    WINML_EXPECT_THROW_SPECIFIC(LearningModelBinding binding(session),
         winrt::hresult_error,
         [](const winrt::hresult_error &e) -> bool
     {
         return e.code() == RO_E_CLOSED;
     });
  }
+
+const LearningModelSesssionAPITestApi& getapi() {
+  static constexpr LearningModelSesssionAPITestApi api =
+  {
+    LearningModelSessionAPITestSetup,
+    LearningModelSessionAPITestGpuSetup,
+    LearningModelSessionAPITestsSkipEdgeCoreSetup,
+    CreateSessionDeviceDefault,
+    CreateSessionDeviceCpu,
+    CreateSessionWithModelLoadedFromStream,
+    CreateSessionDeviceDirectX,
+    CreateSessionDeviceDirectXHighPerformance,
+    CreateSessionDeviceDirectXMinimumPower,
+    AdapterIdAndDevice,
+    EvaluateFeatures,
+    EvaluateFeaturesAsync,
+    EvaluationProperties,
+    CreateSessionWithCastToFloat16InModel,
+    DISABLED_CreateSessionWithFloat16InitializersInModel,
+    EvaluateSessionAndCloseModel,
+    CloseSession,
+  };
+ return api;
+}
\ No newline at end of file
diff --git a/winml/test/api/LearningModelSessionAPITest.h b/winml/test/api/LearningModelSessionAPITest.h
new file mode 100644
index 0000000000000..b98cb56f9fbd7
--- /dev/null
+++ b/winml/test/api/LearningModelSessionAPITest.h
@@ -0,0 +1,44 @@
+#include "test.h"
+
+struct LearningModelSesssionAPITestApi {
+  SetupTest LearningModelSessionAPITestSetup;
+  SetupTest LearningModelSessionAPITestGpuSetup;
+  SetupTest LearningModelSessionAPITestsSkipEdgeCoreSetup;
+  VoidTest CreateSessionDeviceDefault;
+  VoidTest CreateSessionDeviceCpu;
+  VoidTest CreateSessionWithModelLoadedFromStream;
+  VoidTest CreateSessionDeviceDirectX;
+  VoidTest CreateSessionDeviceDirectXHighPerformance;
+  VoidTest CreateSessionDeviceDirectXMinimumPower;
+  VoidTest AdapterIdAndDevice;
+  VoidTest EvaluateFeatures;
+  VoidTest EvaluateFeaturesAsync;
+  VoidTest EvaluationProperties;
+  VoidTest CreateSessionWithCastToFloat16InModel;
+  VoidTest DISABLED_CreateSessionWithFloat16InitializersInModel;
+  VoidTest EvaluateSessionAndCloseModel;
+  VoidTest CloseSession;
+};
+const LearningModelSesssionAPITestApi& getapi();
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelSessionAPITest, LearningModelSessionAPITestSetup)
+WINML_TEST(LearningModelSessionAPITest, CreateSessionDeviceDefault)
+WINML_TEST(LearningModelSessionAPITest,CreateSessionDeviceCpu)
+WINML_TEST(LearningModelSessionAPITest,CreateSessionWithModelLoadedFromStream)
+WINML_TEST(LearningModelSessionAPITest,EvaluateFeatures)
+WINML_TEST(LearningModelSessionAPITest,EvaluateFeaturesAsync)
+WINML_TEST(LearningModelSessionAPITest,EvaluationProperties)
+WINML_TEST(LearningModelSessionAPITest,EvaluateSessionAndCloseModel)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelSessionAPITestGpu, LearningModelSessionAPITestGpuSetup)
+WINML_TEST(LearningModelSessionAPITestGpu, CreateSessionDeviceDirectX)
+WINML_TEST(LearningModelSessionAPITestGpu, CreateSessionDeviceDirectXHighPerformance)
+WINML_TEST(LearningModelSessionAPITestGpu, CreateSessionDeviceDirectXMinimumPower)
+WINML_TEST(LearningModelSessionAPITestGpu, CreateSessionWithCastToFloat16InModel)
+WINML_TEST(LearningModelSessionAPITestGpu, DISABLED_CreateSessionWithFloat16InitializersInModel)
+WINML_TEST_CLASS_END()
+
+WINML_TEST_CLASS_BEGIN_WITH_SETUP(LearningModelSessionAPITestsSkipEdgeCore, LearningModelSessionAPITestsSkipEdgeCoreSetup)
+WINML_TEST(LearningModelSessionAPITestsSkipEdgeCore, AdapterIdAndDevice)
+WINML_TEST_CLASS_END()
\ No newline at end of file
diff --git a/winml/test/common/googleTestMacros.h b/winml/test/common/googleTestMacros.h
index 55fb2f7f740fe..dc38c6402f005 100644
--- a/winml/test/common/googleTestMacros.h
+++ b/winml/test/common/googleTestMacros.h
@@ -30,17 +30,34 @@
 #define GTEST_SKIP GTEST_SKIP_("")
 #endif
 
+#define EXPECT_THROW_SPECIFIC(statement, exception, condition)  \
+    EXPECT_THROW(                                               \
+        try {                                                   \
+            statement;                                          \
+        } catch (const exception& e) {                          \
+            EXPECT_TRUE(condition(e));                          \
+            throw;                                              \
+        }                                                       \
+    , exception);
+
+#ifndef INSTANTIATE_TEST_SUITE_P
+// Use the old name, removed in newer versions of googletest
+#define INSTANTIATE_TEST_SUITE_P INSTANTIATE_TEST_CASE_P
+#endif
+
 #define WINML_SKIP_TEST(message) \
   GTEST_SKIP() << message;
 
 #define WINML_EXPECT_NO_THROW(statement) EXPECT_NO_THROW(statement)
 #define WINML_EXPECT_TRUE(statement) EXPECT_TRUE(statement)
+#define WINML_EXPECT_FALSE(statement) EXPECT_FALSE(statement)
 #define WINML_EXPECT_EQUAL(val1, val2) EXPECT_EQ(val1, val2)
 #define WINML_EXPECT_NOT_EQUAL(val1, val2) EXPECT_NE(val1, val2)
 
 #define WINML_LOG_ERROR(message) \
   ADD_FAILURE() << message
-
+#define WINML_LOG_COMMENT(message)\
+  SCOPED_TRACE(message)
 #define WINML_EXPECT_HRESULT_SUCCEEDED(hresult_expression) EXPECT_HRESULT_SUCCEEDED(hresult_expression)
 #define WINML_EXPECT_HRESULT_FAILED(hresult_expression) EXPECT_HRESULT_FAILED(hresult_expression)
 #define WINML_EXPECT_THROW_SPECIFIC(statement, exception, condition) EXPECT_THROW_SPECIFIC(statement, exception, condition)
diff --git a/winml/test/common/std.h b/winml/test/common/std.h
index 162915abd02fd..d6eab75f645f8 100644
--- a/winml/test/common/std.h
+++ b/winml/test/common/std.h
@@ -29,19 +29,4 @@
 #include "comp_generated/winrt/windows.ai.machinelearning.h"
 
 // WinML
-#include "Windows.AI.MachineLearning.Native.h"
-
-#define EXPECT_THROW_SPECIFIC(statement, exception, condition)  \
-    EXPECT_THROW(                                               \
-        try {                                                   \
-            statement;                                          \
-        } catch (const exception& e) {                          \
-            EXPECT_TRUE(condition(e));                          \
-            throw;                                              \
-        }                                                       \
-    , exception);
-
-#ifndef INSTANTIATE_TEST_SUITE_P
-// Use the old name, removed in newer versions of googletest
-#define INSTANTIATE_TEST_SUITE_P INSTANTIATE_TEST_CASE_P
-#endif
\ No newline at end of file
+#include "Windows.AI.MachineLearning.Native.h"
\ No newline at end of file
diff --git a/winml/test/common/taefTestMacros.h b/winml/test/common/taefTestMacros.h
index 7c7636b7a4dda..06d9048d7c6ff 100644
--- a/winml/test/common/taefTestMacros.h
+++ b/winml/test/common/taefTestMacros.h
@@ -30,11 +30,13 @@ using namespace WEX::TestExecution;
 
 #define WINML_EXPECT_NO_THROW(statement) VERIFY_NO_THROW(statement)
 #define WINML_EXPECT_TRUE(statement) VERIFY_IS_TRUE(statement)
+#define WINML_EXPECT_FALSE(statement) VERIFY_IS_FALSE(statement)
 #define WINML_EXPECT_EQUAL(val1, val2) VERIFY_ARE_EQUAL(val1, val2)
 #define WINML_EXPECT_NOT_EQUAL(val1, val2) VERIFY_ARE_NOT_EQUAL(val1, val2)
 #define WINML_LOG_ERROR(message) \
   VERIFY_FAIL(std::wstring_convert<std::codecvt_utf8<wchar_t>>().from_bytes(message).c_str())
-
+#define WINML_LOG_COMMENT(message)\
+  WEX::Logging::Log::Comment(std::wstring_convert<std::codecvt_utf8<wchar_t>>().from_bytes(message).c_str())
 #define WINML_EXPECT_HRESULT_SUCCEEDED(hresult_expression) VERIFY_SUCCEEDED(hresult_expression)
 #define WINML_EXPECT_THROW_SPECIFIC(statement, exception, condition) VERIFY_THROWS_SPECIFIC(statement, exception, condition)
 #define WINML_EXPECT_HRESULT_FAILED(hresult_expression) VERIFY_FAILED(hresult_expression)

From 39c9f98a5d85ebe6458fc2b01e75f1dfbb95e3ec Mon Sep 17 00:00:00 2001
From: Ryan Lai <ryalai96@gmail.com>
Date: Wed, 15 Jan 2020 12:45:34 -0800
Subject: [PATCH 64/75] Fix test build break since winml_lib_api needs to be
 statically linked to tests since winmlp::learningmodeldevice::iscpu() is
 being used in devicehelpers.cpp (#2837)

---
 cmake/winml_unittests.cmake | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/cmake/winml_unittests.cmake b/cmake/winml_unittests.cmake
index 4bdbb34432466..bd91611a6d17c 100644
--- a/cmake/winml_unittests.cmake
+++ b/cmake/winml_unittests.cmake
@@ -44,7 +44,7 @@ function(add_winml_test)
   if (_UT_DEPENDS)
     add_dependencies(${_UT_TARGET} ${_UT_DEPENDS})
   endif()
-  target_link_libraries(${_UT_TARGET} PRIVATE ${_UT_LIBS} gtest windowsapp winml_lib_image ${onnxruntime_EXTERNAL_LIBRARIES} winml_lib_telemetry)
+  target_link_libraries(${_UT_TARGET} PRIVATE ${_UT_LIBS} gtest windowsapp winml_lib_image ${onnxruntime_EXTERNAL_LIBRARIES} winml_lib_telemetry winml_lib_api onnxruntime)
 
   add_test(NAME ${_UT_TARGET}
     COMMAND ${_UT_TARGET}

From 408e600bf3c8ce5ce3754f5afa7d8aeaeee14429 Mon Sep 17 00:00:00 2001
From: Tiago Koji Castro Shibata <ticastro@microsoft.com>
Date: Wed, 15 Jan 2020 15:35:55 -0800
Subject: [PATCH 65/75] Enforce WINML_TEST_CLASS_BEGIN_* matches w/ a
 WINML_TEST_CLASS_END (#2841)

---
 winml/test/common/googleTestMacros.h | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/winml/test/common/googleTestMacros.h b/winml/test/common/googleTestMacros.h
index dc38c6402f005..f0507435b95c8 100644
--- a/winml/test/common/googleTestMacros.h
+++ b/winml/test/common/googleTestMacros.h
@@ -10,18 +10,20 @@
   }
 
 #define WINML_TEST_CLASS_BEGIN_NO_SETUP(test_class_name) \
-  class test_class_name : public ::testing::Test {       \
-  };
+  namespace {                                            \
+    class test_class_name : public ::testing::Test {     \
+    };
 
 #define WINML_TEST_CLASS_BEGIN_WITH_SETUP(test_class_name, setup_method) \
-  class test_class_name : public ::testing::Test {                       \
-   protected:                                                            \
-    void SetUp() override {                                              \
-      getapi().setup_method();                                           \
-    }                                                                    \
-  };
+  namespace {                                                            \
+    class test_class_name : public ::testing::Test {                     \
+    protected:                                                           \
+      void SetUp() override {                                            \
+        getapi().setup_method();                                         \
+      }                                                                  \
+    };
 
-#define WINML_TEST_CLASS_END()
+#define WINML_TEST_CLASS_END() }
 
 // For old versions of gtest without GTEST_SKIP, stream the message and return success instead
 #ifndef GTEST_SKIP
@@ -77,4 +79,4 @@
   if (auto isEdgeCore = RuntimeParameters::Parameters.find("EdgeCore");                 \
       isEdgeCore != RuntimeParameters::Parameters.end() && isEdgeCore->second != "0") { \
     WINML_SKIP_TEST("Test can't be run in EdgeCore");                                   \
-  }
\ No newline at end of file
+  }

From 5db8543018a2c2796defa714111633c32d4a2a33 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Wed, 15 Jan 2020 16:01:11 -0800
Subject: [PATCH 66/75] update optimization doc for BERT related fusions 
 (#2819)

* Add bert related transformers to doc
* Add execution provider and comment for bert optimizations
* Add comment about accuracy impact of approximation
---
 docs/ONNX_Runtime_Graph_Optimizations.md | 23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)

diff --git a/docs/ONNX_Runtime_Graph_Optimizations.md b/docs/ONNX_Runtime_Graph_Optimizations.md
index 2d124420db980..1002b41d68c64 100644
--- a/docs/ONNX_Runtime_Graph_Optimizations.md
+++ b/docs/ONNX_Runtime_Graph_Optimizations.md
@@ -34,15 +34,26 @@ These are semantics-preserving graph rewrites which remove redundant nodes and r
   * Conv Mul Fusion
   * Conv BatchNorm Fusion
   * Relu Clip Fusion
+  * Reshape Fusion
 
 ### Extended Graph Optimizations
 
-These optimizations include complex node fusions. They are run after graph partitioning and are only applied to the nodes assigned to the CPU execution provider. Available extended graph optimizations are as follows:
-
-* GEMM Activation Fusion
-* Matmul Add Fusion
-* Conv Activation Fusion
-* GELU Fusion
+These optimizations include complex node fusions. They are run after graph partitioning and are only applied to the nodes assigned to the CPU or CUDA execution provider. Available extended graph optimizations are as follows:
+
+| Optimization                    | Execution Provider | Comment                                                                     |
+|---------------------------------|--------------------|-----------------------------------------------------------------------------|
+| GEMM Activation Fusion          | cpu                |                                                                             |
+| Matmul Add Fusion               | cpu                |                                                                             |
+| Conv Activation Fusion          | cpu                |                                                                             |
+| GELU Fusion                     | cpu or cuda        |                                                                             |
+| Layer Normalization Fusion      | cpu or cuda        |                                                                             |
+| BERT Embedding Layer Fusion     | cpu or cuda        | Fuse BERT embedding layer, layer normalization and attention mask length    |
+| Attention Fusion                | cpu or cuda        | Attention mask has approximation in cuda execution provider                 |
+| Skip Layer Normalization Fusion | cpu or cuda        | Fuse bias of fully connected layer, skip connection and layer normalization |
+| Bias GELU Fusion                | cpu or cuda        | Fuse bias of fully connected layer and GELU activation                      |
+| GELU Approximation              | cuda               | Erf is approximated by a formula using tanh function                        |
+
+To optimize inference performance of BERT model, approximation is used in GELU approximation and Attention fusion for cuda execution provider. There might be slight difference in result. The impact on accuracy could be neglected based on our evaluation: F1 score for a BERT model on SQuAD v1.1 is almost same (87.05 vs 87.03).
 
 ### Layout Optimizations
 

From 70523a3df53f5fe37ac1dd1252fe47a0681d0d45 Mon Sep 17 00:00:00 2001
From: Ryan Lai <rylai@microsoft.com>
Date: Wed, 15 Jan 2020 17:07:22 -0800
Subject: [PATCH 67/75] Fix warnings that cause build to fail

---
 .../core/providers/dml/OperatorAuthorHelper/OperatorHelper.h | 5 +++--
 winml/test/common/protobufHelpers.cpp                        | 4 ++--
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h
index aeec2deeec36c..c944c1b59b27e 100644
--- a/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h
+++ b/onnxruntime/core/providers/dml/OperatorAuthorHelper/OperatorHelper.h
@@ -437,6 +437,7 @@ class GemmHelper
     template<typename Info_t, typename Shape_t>
     GemmHelper(const Info_t& info, const Shape_t& shape)
     {
+        ORT_UNUSED_PARAMETER(shape);
         m_transA = info.GetOptionalAttribute<int>(AttrName::TransA, 0);
         m_transB = info.GetOptionalAttribute<int>(AttrName::TransB, 0);
         m_broadcast = info.GetOptionalAttribute<int>(AttrName::Broadcast, 0);
@@ -528,7 +529,7 @@ class SliceHelperBase
             ends.push_back(gsl::narrow_cast<int32_t>(endsData[i]));
         }
         uint32_t inputCount = operatorInfo.GetInputCount();
-        if (operatorInfo.GetInputCount() > 3)
+        if (inputCount > 3)
         {
             MLOperatorTensor axesTensor = operatorInfo.GetConstantInputTensor(3);
             const std::vector<uint32_t>& axesTensorDimensions = axesTensor.GetShape();
@@ -540,7 +541,7 @@ class SliceHelperBase
             }
         }
 
-        if (operatorInfo.GetInputCount() > 4)
+        if (inputCount > 4)
         {
             MLOperatorTensor stepsTensor = operatorInfo.GetConstantInputTensor(4);
             const std::vector<uint32_t>& stepsTensorDimensions = stepsTensor.GetShape();
diff --git a/winml/test/common/protobufHelpers.cpp b/winml/test/common/protobufHelpers.cpp
index 0f3c22fb1c36b..dbd440b9ab71b 100644
--- a/winml/test/common/protobufHelpers.cpp
+++ b/winml/test/common/protobufHelpers.cpp
@@ -323,9 +323,9 @@ winrt::Windows::AI::MachineLearning::LearningModel ProtobufHelpers::CreateModel(
     DataWriter m_dataWriter;
   };
 
-  auto size = model.ByteSize();
+  auto size = model.ByteSizeLong();
   auto raw_array = std::unique_ptr<char[]>(new char[size]);
-  model.SerializeToArray(raw_array.get(), size);
+  model.SerializeToArray(raw_array.get(), static_cast<int>(size));
 
   BufferStreamAdapter buffer;
   std::ostream os(&buffer);

From 928b6bb21010b06c446248fa89ee95f11e61dded Mon Sep 17 00:00:00 2001
From: Tracy Sharpe <42477615+tracysh@users.noreply.github.com>
Date: Wed, 15 Jan 2020 19:25:40 -0800
Subject: [PATCH 68/75] MLAS: enable threading for quantized GEMMs (#2844)

---
 onnxruntime/contrib_ops/cpu/nchwc_ops.cc      |   1 -
 .../mlas/lib/amd64/QgemmU8S8KernelAvx2.asm    |  68 +--
 onnxruntime/core/mlas/lib/mlasi.h             |  42 ++
 onnxruntime/core/mlas/lib/qgemm.cpp           | 524 ++++++++++++++++--
 onnxruntime/core/mlas/lib/snchwc.cpp          |  30 +-
 onnxruntime/core/mlas/lib/threading.cpp       |   1 -
 .../mlas/lib/x86_64/QgemmU8S8KernelAvx2.S     |  68 +--
 onnxruntime/core/providers/cpu/math/gemm.h    |   5 +-
 onnxruntime/core/providers/cpu/math/matmul.cc |   3 +-
 .../core/providers/cpu/math/matmul_integer.cc |   1 -
 onnxruntime/core/providers/cpu/nn/conv.cc     |  12 +-
 .../core/providers/cpu/nn/conv_integer.cc     |   3 +-
 .../core/providers/cpu/nn/conv_transpose.cc   |   9 +-
 onnxruntime/core/providers/cpu/nn/pool.cc     |   1 -
 14 files changed, 602 insertions(+), 166 deletions(-)

diff --git a/onnxruntime/contrib_ops/cpu/nchwc_ops.cc b/onnxruntime/contrib_ops/cpu/nchwc_ops.cc
index 325217a1f8e10..14e0a83384e7a 100644
--- a/onnxruntime/contrib_ops/cpu/nchwc_ops.cc
+++ b/onnxruntime/contrib_ops/cpu/nchwc_ops.cc
@@ -1,7 +1,6 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
-#include "core/framework/op_kernel_context_internal.h"
 #include "nchwc_ops.h"
 #include "core/mlas/inc/mlas.h"
 
diff --git a/onnxruntime/core/mlas/lib/amd64/QgemmU8S8KernelAvx2.asm b/onnxruntime/core/mlas/lib/amd64/QgemmU8S8KernelAvx2.asm
index f701fb623facb..a4d9ec8468a8f 100644
--- a/onnxruntime/core/mlas/lib/amd64/QgemmU8S8KernelAvx2.asm
+++ b/onnxruntime/core/mlas/lib/amd64/QgemmU8S8KernelAvx2.asm
@@ -170,6 +170,8 @@ ProcessNextRowM4:
         vpxor   xmm1,xmm1,xmm1
         vpxor   xmm2,xmm2,xmm2
         vpxor   xmm3,xmm3,xmm3
+        lea     r13,[r8+r8*2]               ; compute ldb * 3
+        lea     rax,[r11+r11*2]             ; compute output stride * 3
         mov     rdx,rsi
         mov     rcx,rdi
         lea     rsi,[rsi+r8*4]              ; advance next matrix A by 4 rows
@@ -179,16 +181,14 @@ ProcessNextRowM4:
         jb      ProcessRemainingColumnsM4
 
 ProcessNextColumnLoopM4:
-        lea     rax,[rdx+r8*2]              ; compute matrix A plus 2 rows
         vmovdqu ymm4,YMMWORD PTR [rdx]
         vmovdqu ymm5,YMMWORD PTR [rdx+r8]
-        vmovdqu ymm6,YMMWORD PTR [rax]
-        vmovdqu ymm7,YMMWORD PTR [rax+r8]
-        lea     rax,[rcx+r11*2]             ; compute matrix D plus 2 rows
+        vmovdqu ymm6,YMMWORD PTR [rdx+r8*2]
+        vmovdqu ymm7,YMMWORD PTR [rdx+r13]
         vmovdqu YMMWORD PTR [rcx],ymm4
         vmovdqu YMMWORD PTR [rcx+r11],ymm5
-        vmovdqu YMMWORD PTR [rax],ymm6
-        vmovdqu YMMWORD PTR [rax+r11],ymm7
+        vmovdqu YMMWORD PTR [rcx+r11*2],ymm6
+        vmovdqu YMMWORD PTR [rcx+rax],ymm7
         vpmaddubsw ymm4,ymm4,ymm9           ; horizontal byte+byte=word per row
         vpaddw  ymm0,ymm0,ymm4              ; add words to row accumulators
         vpmaddubsw ymm5,ymm5,ymm9
@@ -207,16 +207,14 @@ ProcessRemainingColumnsM4:
         jz      ReduceRowSumVectorM4
         test    bl,16                       ; (CountK & 16) != 0?
         jz      CopyRemainingCountKLessThan16M4
-        lea     rax,[rdx+r8*2]              ; compute matrix A plus 2 rows
         vmovdqu xmm4,XMMWORD PTR [rdx]
         vmovdqu xmm5,XMMWORD PTR [rdx+r8]
-        vmovdqu xmm6,XMMWORD PTR [rax]
-        vmovdqu xmm7,XMMWORD PTR [rax+r8]
-        lea     rax,[rcx+r11*2]             ; compute matrix D plus 2 rows
+        vmovdqu xmm6,XMMWORD PTR [rdx+r8*2]
+        vmovdqu xmm7,XMMWORD PTR [rdx+r13]
         vmovdqu XMMWORD PTR [rcx],xmm4
         vmovdqu XMMWORD PTR [rcx+r11],xmm5
-        vmovdqu XMMWORD PTR [rax],xmm6
-        vmovdqu XMMWORD PTR [rax+r11],xmm7
+        vmovdqu XMMWORD PTR [rcx+r11*2],xmm6
+        vmovdqu XMMWORD PTR [rcx+rax],xmm7
         vpmaddubsw xmm4,xmm4,xmm9           ; horizontal byte+byte=word per row
         vpaddw  ymm0,ymm0,ymm4              ; add words to row accumulators
         vpmaddubsw xmm5,xmm5,xmm9
@@ -239,14 +237,13 @@ CopyRemainingCountKLessThan16M4:
         mov     rbp,rsp                     ; GemmU8S8CopyPackAFrame.PaddedMatrixAData
         test    bl,8                        ; (CountK & 8) != 0?
         jz      CopyRemainingCountKLessThan8M4
-        lea     r13,[rdx+r8*2]              ; compute matrix A plus 2 rows
         mov     rax,QWORD PTR [rdx]
         mov     QWORD PTR [rbp],rax
         mov     rax,QWORD PTR [rdx+r8]
         mov     QWORD PTR [rbp+16],rax
-        mov     rax,QWORD PTR [r13]
+        mov     rax,QWORD PTR [rdx+r8*2]
         mov     QWORD PTR [rbp+32],rax
-        mov     rax,QWORD PTR [r13+r8]
+        mov     rax,QWORD PTR [rdx+r13]
         mov     QWORD PTR [rbp+48],rax
         add     rdx,8
         add     rbp,8                       ; advance padded buffer destination
@@ -254,14 +251,13 @@ CopyRemainingCountKLessThan16M4:
 CopyRemainingCountKLessThan8M4:
         test    bl,4                        ; (CountK & 4) != 0?
         jz      CopyRemainingCountKLessThan4M4
-        lea     r13,[rdx+r8*2]              ; compute matrix A plus 2 rows
         mov     eax,DWORD PTR [rdx]
         mov     DWORD PTR [rbp],eax
         mov     eax,DWORD PTR [rdx+r8]
         mov     DWORD PTR [rbp+16],eax
-        mov     eax,DWORD PTR [r13]
+        mov     eax,DWORD PTR [rdx+r8*2]
         mov     DWORD PTR [rbp+32],eax
-        mov     eax,DWORD PTR [r13+r8]
+        mov     eax,DWORD PTR [rdx+r13]
         mov     DWORD PTR [rbp+48],eax
         add     rdx,4
         add     rbp,4                       ; advance padded buffer destination
@@ -269,14 +265,13 @@ CopyRemainingCountKLessThan8M4:
 CopyRemainingCountKLessThan4M4:
         test    bl,2                        ; (CountK & 2) != 0?
         jz      CopyRemainingCountKLessThan2M4
-        lea     r13,[rdx+r8*2]              ; compute matrix A plus 2 rows
         movzx   eax,WORD PTR [rdx]
         mov     WORD PTR [rbp],ax
         movzx   eax,WORD PTR [rdx+r8]
         mov     WORD PTR [rbp+16],ax
-        movzx   eax,WORD PTR [r13]
+        movzx   eax,WORD PTR [rdx+r8*2]
         mov     WORD PTR [rbp+32],ax
-        movzx   eax,WORD PTR [r13+r8]
+        movzx   eax,WORD PTR [rdx+r13]
         mov     WORD PTR [rbp+48],ax
         add     rdx,2
         add     rbp,2                       ; advance padded buffer destination
@@ -284,14 +279,13 @@ CopyRemainingCountKLessThan4M4:
 CopyRemainingCountKLessThan2M4:
         test    bl,1                        ; (CountK & 1) != 0?
         jz      ProcessPaddedMatrixADataM4
-        lea     r13,[rdx+r8*2]              ; compute matrix A plus 2 rows
         movzx   eax,BYTE PTR [rdx]
         mov     BYTE PTR [rbp],al
         movzx   eax,BYTE PTR [rdx+r8]
         mov     BYTE PTR [rbp+16],al
-        movzx   eax,BYTE PTR [r13]
+        movzx   eax,BYTE PTR [rdx+r8*2]
         mov     BYTE PTR [rbp+32],al
-        movzx   eax,BYTE PTR [r13+r8]
+        movzx   eax,BYTE PTR [rdx+r13]
         mov     BYTE PTR [rbp+48],al
 
 ;
@@ -507,6 +501,7 @@ ExitRoutine:
         END_PROLOGUE
 
         mov     rsi,rdx
+        lea     rdi,[r8+r8*2]               ; compute ldb * 3
         mov     r10,GemmU8S8CopyPackBFrame.CountK[rsp]
         mov     r11,GemmU8S8CopyPackBFrame.ColumnSumVector[rsp]
         vpbroadcastw ymm7,WORD PTR GemmU8S8CopyPackBFrame.offa[rsp]
@@ -532,11 +527,10 @@ ProcessNextColumnN16:
         jb      ProcessRemainingRowsN16
 
 ProcessNextRowLoopN16:
-        lea     rax,[rdx+r8*2]              ; compute matrix B plus 2 rows
         vmovdqu xmm2,XMMWORD PTR [rdx]      ; load 4 rows
         vmovdqu xmm3,XMMWORD PTR [rdx+r8]
-        vmovdqu xmm4,XMMWORD PTR [rax]
-        vmovdqu xmm5,XMMWORD PTR [rax+r8]
+        vmovdqu xmm4,XMMWORD PTR [rdx+r8*2]
+        vmovdqu xmm5,XMMWORD PTR [rdx+rdi]
         lea     rdx,[rdx+r8*4]              ; advance matrix B by 4 rows
 
 InterleaveRowDataN16:
@@ -630,14 +624,13 @@ ProcessNextRowLoopNUnaligned:
         mov     rbp,rsp                     ; GemmU8S8CopyPackBFrame.PaddedMatrixBData
         test    r9b,8                       ; (CountN & 8) != 0?
         jz      CopyRemainingCountNLessThan8K4
-        lea     rdi,[rdx+r8*2]              ; compute matrix B plus 2 rows
         mov     rax,QWORD PTR [rdx]
         mov     QWORD PTR [rbp],rax
         mov     rax,QWORD PTR [rdx+r8]
         mov     QWORD PTR [rbp+16],rax
-        mov     rax,QWORD PTR [rdi]
+        mov     rax,QWORD PTR [rdx+r8*2]
         mov     QWORD PTR [rbp+32],rax
-        mov     rax,QWORD PTR [rdi+r8]
+        mov     rax,QWORD PTR [rdx+rdi]
         mov     QWORD PTR [rbp+48],rax
         add     rdx,8                       ; advance matrix B
         add     rbp,8                       ; advance padded buffer destination
@@ -645,14 +638,13 @@ ProcessNextRowLoopNUnaligned:
 CopyRemainingCountNLessThan8K4:
         test    r9b,4                       ; (CountN & 4) != 0?
         jz      CopyRemainingCountNLessThan4K4
-        lea     rdi,[rdx+r8*2]              ; compute matrix B plus 2 rows
         mov     eax,DWORD PTR [rdx]
         mov     DWORD PTR [rbp],eax
         mov     eax,DWORD PTR [rdx+r8]
         mov     DWORD PTR [rbp+16],eax
-        mov     eax,DWORD PTR [rdi]
+        mov     eax,DWORD PTR [rdx+r8*2]
         mov     DWORD PTR [rbp+32],eax
-        mov     eax,DWORD PTR [rdi+r8]
+        mov     eax,DWORD PTR [rdx+rdi]
         mov     DWORD PTR [rbp+48],eax
         add     rdx,4                       ; advance matrix B
         add     rbp,4                       ; advance padded buffer destination
@@ -660,14 +652,13 @@ CopyRemainingCountNLessThan8K4:
 CopyRemainingCountNLessThan4K4:
         test    r9b,2                       ; (CountN & 2) != 0?
         jz      CopyRemainingCountNLessThan2K4
-        lea     rdi,[rdx+r8*2]              ; compute matrix B plus 2 rows
         movzx   eax,WORD PTR [rdx]
         mov     WORD PTR [rbp],ax
         movzx   eax,WORD PTR [rdx+r8]
         mov     WORD PTR [rbp+16],ax
-        movzx   eax,WORD PTR [rdi]
+        movzx   eax,WORD PTR [rdx+r8*2]
         mov     WORD PTR [rbp+32],ax
-        movzx   eax,WORD PTR [rdi+r8]
+        movzx   eax,WORD PTR [rdx+rdi]
         mov     WORD PTR [rbp+48],ax
         add     rdx,2                       ; advance matrix B
         add     rbp,2                       ; advance padded buffer destination
@@ -675,14 +666,13 @@ CopyRemainingCountNLessThan4K4:
 CopyRemainingCountNLessThan2K4:
         test    r9b,1                       ; (CountN & 1) != 0?
         jz      ProcessPaddedMatrixBData
-        lea     rdi,[rdx+r8*2]              ; compute matrix B plus 2 rows
         movzx   eax,BYTE PTR [rdx]
         mov     BYTE PTR [rbp],al
         movzx   eax,BYTE PTR [rdx+r8]
         mov     BYTE PTR [rbp+16],al
-        movzx   eax,BYTE PTR [rdi]
+        movzx   eax,BYTE PTR [rdx+r8*2]
         mov     BYTE PTR [rbp+32],al
-        movzx   eax,BYTE PTR [rdi+r8]
+        movzx   eax,BYTE PTR [rdx+rdi]
         mov     BYTE PTR [rbp+48],al
 
 ProcessPaddedMatrixBData:
diff --git a/onnxruntime/core/mlas/lib/mlasi.h b/onnxruntime/core/mlas/lib/mlasi.h
index ff0b038cf200a..12ed28672e5c2 100644
--- a/onnxruntime/core/mlas/lib/mlasi.h
+++ b/onnxruntime/core/mlas/lib/mlasi.h
@@ -131,6 +131,7 @@ Module Name:
 
 #define MLAS_SGEMM_STRIDEN_THREAD_ALIGN             16
 #define MLAS_DGEMM_STRIDEN_THREAD_ALIGN             8
+#define MLAS_QGEMM_STRIDEN_THREAD_ALIGN             16
 
 //
 // Define the prototypes of the platform optimized routines.
@@ -333,6 +334,24 @@ size_t
 
 typedef MLAS_GEMM_U8U8_KERNEL* PMLAS_GEMM_U8U8_KERNEL;
 
+typedef
+void
+(MLASCALL MLAS_GEMM_X8X8_OPERATION)(
+    size_t M,
+    size_t N,
+    size_t K,
+    const uint8_t* A,
+    size_t lda,
+    int16_t offa,
+    const uint8_t* B,
+    size_t ldb,
+    int16_t offb,
+    int32_t* C,
+    size_t ldc
+    );
+
+typedef MLAS_GEMM_X8X8_OPERATION* PMLAS_GEMM_X8X8_OPERATION;
+
 typedef
 void
 (MLASCALL MLAS_CONV_FLOAT_KERNEL)(
@@ -557,6 +576,7 @@ extern "C" {
 
 #define MLAS_SGEMM_THREAD_COMPLEXITY                (64 * 1024)
 #define MLAS_DGEMM_THREAD_COMPLEXITY                (64 * 1024)
+#define MLAS_QGEMM_THREAD_COMPLEXITY                (64 * 1024)
 
 //
 // Single-threaded single precision matrix/matrix multiply operation.
@@ -660,6 +680,28 @@ MlasGetMaximumThreadCount(
 #endif
 }
 
+inline
+void
+MlasPartitionWork(
+    int32_t ThreadId,
+    int32_t ThreadCount,
+    size_t TotalWork,
+    size_t* WorkIndex,
+    size_t* WorkRemaining
+    )
+{
+    const size_t WorkPerThread = TotalWork / ThreadCount;
+    const size_t WorkPerThreadExtra = TotalWork % ThreadCount;
+
+    if (uint32_t(ThreadId) < WorkPerThreadExtra) {
+        *WorkIndex = (WorkPerThread + 1) * ThreadId;
+        *WorkRemaining = WorkPerThread + 1;
+    } else {
+        *WorkIndex = WorkPerThread * ThreadId + WorkPerThreadExtra;
+        *WorkRemaining = WorkPerThread;
+    }
+}
+
 //
 // Define the missing ARM64 NEON intrinsic macros from arm64_neon.h that enable
 // cross-compiler support.
diff --git a/onnxruntime/core/mlas/lib/qgemm.cpp b/onnxruntime/core/mlas/lib/qgemm.cpp
index 776107fc5c97a..377628c3564ca 100644
--- a/onnxruntime/core/mlas/lib/qgemm.cpp
+++ b/onnxruntime/core/mlas/lib/qgemm.cpp
@@ -21,13 +21,33 @@ Module Name:
 // Define the default strides to step through slices of the input matrices.
 //
 
-#define MLAS_GEMM_U8S8_STRIDEM              24
-#define MLAS_GEMM_U8S8_STRIDEN              256
-#define MLAS_GEMM_U8S8_STRIDEK              128
+#define MLAS_GEMM_X8X8_STRIDEM              24
+#define MLAS_GEMM_X8X8_STRIDEN              256
+#define MLAS_GEMM_X8X8_STRIDEK              128
 
-#define MLAS_GEMM_U8U8_STRIDEM              24
-#define MLAS_GEMM_U8U8_STRIDEN              256
-#define MLAS_GEMM_U8U8_STRIDEK              128
+//
+// Define the parameters to execute segments of a QGEMM operation on worker
+// threads.
+//
+
+struct MLAS_GEMM_X8X8_WORK_BLOCK {
+    PMLAS_GEMM_X8X8_OPERATION GemmX8X8Operation;
+    size_t M;
+    size_t N;
+    size_t K;
+    const uint8_t* A;
+    size_t lda;
+    const uint8_t* B;
+    size_t ldb;
+    int32_t* C;
+    size_t ldc;
+    int32_t ThreadCountM;
+    int32_t ThreadCountN;
+    size_t StrideM;
+    size_t StrideN;
+    int16_t offa;
+    int16_t offb;
+};
 
 #ifdef MLAS_TARGET_AMD64_IX86
 
@@ -1102,45 +1122,83 @@ Return Value:
 
 void
 MLASCALL
-MlasGemm(
+MlasGemmU8S8Operation(
     size_t M,
     size_t N,
     size_t K,
     const uint8_t* A,
     size_t lda,
-    uint8_t offa,
-    const int8_t* B,
+    int16_t offa,
+    const uint8_t* B,
     size_t ldb,
-    int8_t offb,
+    int16_t offb,
     int32_t* C,
-    size_t ldc,
-    MLAS_THREADPOOL* ThreadPool
+    size_t ldc
     )
-{
-    MLAS_DECLSPEC_ALIGN(uint8_t PanelA[MLAS_GEMM_U8S8_STRIDEM * MLAS_GEMM_U8S8_STRIDEK], 64);
-    MLAS_DECLSPEC_ALIGN(int8_t PanelB[MLAS_GEMM_U8S8_STRIDEN * MLAS_GEMM_U8S8_STRIDEK], 64);
+/*++
+
+Routine Description:
+
+    This module implements the quantized integer matrix/matrix multiply
+    operation (QGEMM).
+
+Arguments:
+
+    M - Supplies the number of rows of matrix A and matrix C.
+
+    N - Supplies the number of columns of matrix B and matrix C.
+
+    K - Supplies the number of columns of matrix A and the number of rows of
+        matrix B.
+
+    A - Supplies the address of matrix A.
+
+    lda - Supplies the first dimension of matrix A.
+
+    offa - Supplies the zero point offset of matrix A.
+
+    B - Supplies the address of matrix B.
+
+    ldb - Supplies the first dimension of matrix B.
+
+    offb - Supplies the zero point offset of matrix B.
+
+    C - Supplies the address of matrix C.
+
+    ldc - Supplies the first dimension of matrix C.
+
+Return Value:
+
+    None.
 
-    MLAS_DECLSPEC_ALIGN(int32_t RowSumVector[MLAS_GEMM_U8S8_STRIDEM], 16);
-    MLAS_DECLSPEC_ALIGN(int32_t ColumnSumVector[MLAS_GEMM_U8S8_STRIDEN], 16);
+--*/
+{
+    MLAS_DECLSPEC_ALIGN(uint8_t PanelA[MLAS_GEMM_X8X8_STRIDEM * MLAS_GEMM_X8X8_STRIDEK], 64);
+    MLAS_DECLSPEC_ALIGN(int8_t PanelB[MLAS_GEMM_X8X8_STRIDEN * MLAS_GEMM_X8X8_STRIDEK], 64);
 
-    size_t StrideM = MLAS_GEMM_U8S8_STRIDEM;
-    size_t StrideN = MLAS_GEMM_U8S8_STRIDEN;
-    size_t StrideK = MLAS_GEMM_U8S8_STRIDEK;
+    MLAS_DECLSPEC_ALIGN(int32_t RowSumVector[MLAS_GEMM_X8X8_STRIDEM], 16);
+    MLAS_DECLSPEC_ALIGN(int32_t ColumnSumVector[MLAS_GEMM_X8X8_STRIDEN], 16);
 
-    MLAS_UNREFERENCED_PARAMETER(ThreadPool);
+    size_t StrideM = MLAS_GEMM_X8X8_STRIDEM;
+    size_t StrideN = MLAS_GEMM_X8X8_STRIDEN;
+    size_t StrideK = MLAS_GEMM_X8X8_STRIDEK;
 
 #if defined(MLAS_TARGET_AMD64)
 
     if (M == 1 && offa == 0 && offb == 0) {
 
         if (MlasPlatform.GemvU8S8Kernel != nullptr) {
-            MlasPlatform.GemvU8S8Kernel(A, B, C, K, N, ldb);
+            MlasPlatform.GemvU8S8Kernel(A, (const int8_t*)B, C, K, N, ldb);
             return;
         }
     }
 
 #endif
 
+    //
+    // Step through each slice of matrix B along the K dimension.
+    //
+
     size_t CountK;
 
     for (size_t k = 0; k < K; k += CountK) {
@@ -1151,6 +1209,10 @@ MlasGemm(
             CountK = K - k;
         }
 
+        //
+        // Step through each slice of matrix B along the N dimension.
+        //
+
         size_t CountN;
 
         for (size_t n = 0; n < N; n += CountN) {
@@ -1161,7 +1223,10 @@ MlasGemm(
                 CountN = N - n;
             }
 
-            MlasPlatform.GemmU8S8CopyPackBRoutine(PanelB, B + n + k * ldb, ldb, CountN, CountK, ColumnSumVector, -int16_t(offa));
+            const int8_t* b = (const int8_t*)B + n + k * ldb;
+
+            MlasPlatform.GemmU8S8CopyPackBRoutine(PanelB, b, ldb, CountN,
+                CountK, ColumnSumVector, -int16_t(offa));
 
             size_t CountM;
 
@@ -1173,7 +1238,8 @@ MlasGemm(
                     CountM = M - m;
                 }
 
-                MlasPlatform.GemmU8S8CopyPackARoutine(PanelA, A + k + m * lda, lda, CountM, CountK, RowSumVector, -int16_t(offb));
+                MlasPlatform.GemmU8S8CopyPackARoutine(PanelA, A + k + m * lda,
+                    lda, CountM, CountK, RowSumVector, -int16_t(offb));
 
                 uint8_t* pa = PanelA;
                 int32_t* c = C + n + m * ldc;
@@ -1187,7 +1253,9 @@ MlasGemm(
 
                 while (RowsRemaining > 0) {
 
-                    RowsHandled = MlasPlatform.GemmU8S8Kernel(pa, PanelB, c, QuadCountK, RowsRemaining, CountN, ldc, RowSums, ColumnSumVector, int32_t(CountK) * offa * offb, k == 0);
+                    RowsHandled = MlasPlatform.GemmU8S8Kernel(pa, PanelB, c,
+                        QuadCountK, RowsRemaining, CountN, ldc, RowSums,
+                        ColumnSumVector, int32_t(CountK) * offa * offb, k == 0);
 
                     RowsRemaining -= RowsHandled;
                     c += ldc * RowsHandled;
@@ -1201,32 +1269,70 @@ MlasGemm(
 
 void
 MLASCALL
-MlasGemm(
+MlasGemmU8U8Operation(
     size_t M,
     size_t N,
     size_t K,
     const uint8_t* A,
     size_t lda,
-    uint8_t offa,
+    int16_t offa,
     const uint8_t* B,
     size_t ldb,
-    uint8_t offb,
+    int16_t offb,
     int32_t* C,
-    size_t ldc,
-    MLAS_THREADPOOL* ThreadPool
+    size_t ldc
     )
+/*++
+
+Routine Description:
+
+    This module implements the quantized integer matrix/matrix multiply
+    operation (QGEMM).
+
+Arguments:
+
+    M - Supplies the number of rows of matrix A and matrix C.
+
+    N - Supplies the number of columns of matrix B and matrix C.
+
+    K - Supplies the number of columns of matrix A and the number of rows of
+        matrix B.
+
+    A - Supplies the address of matrix A.
+
+    lda - Supplies the first dimension of matrix A.
+
+    offa - Supplies the zero point offset of matrix A.
+
+    B - Supplies the address of matrix B.
+
+    ldb - Supplies the first dimension of matrix B.
+
+    offb - Supplies the zero point offset of matrix B.
+
+    C - Supplies the address of matrix C.
+
+    ldc - Supplies the first dimension of matrix C.
+
+Return Value:
+
+    None.
+
+--*/
 {
-    MLAS_DECLSPEC_ALIGN(int16_t PanelA[MLAS_GEMM_U8U8_STRIDEM * MLAS_GEMM_U8U8_STRIDEK], 64);
-    MLAS_DECLSPEC_ALIGN(uint8_t PanelB[MLAS_GEMM_U8U8_STRIDEN * MLAS_GEMM_U8U8_STRIDEK], 64);
+    MLAS_DECLSPEC_ALIGN(int16_t PanelA[MLAS_GEMM_X8X8_STRIDEM * MLAS_GEMM_X8X8_STRIDEK], 64);
+    MLAS_DECLSPEC_ALIGN(uint8_t PanelB[MLAS_GEMM_X8X8_STRIDEN * MLAS_GEMM_X8X8_STRIDEK], 64);
 
-    MLAS_DECLSPEC_ALIGN(int32_t RowSumVector[MLAS_GEMM_U8U8_STRIDEM], 16);
-    MLAS_DECLSPEC_ALIGN(int32_t ColumnSumVector[MLAS_GEMM_U8U8_STRIDEN], 16);
+    MLAS_DECLSPEC_ALIGN(int32_t RowSumVector[MLAS_GEMM_X8X8_STRIDEM], 16);
+    MLAS_DECLSPEC_ALIGN(int32_t ColumnSumVector[MLAS_GEMM_X8X8_STRIDEN], 16);
 
-    size_t StrideM = MLAS_GEMM_U8U8_STRIDEM;
-    size_t StrideN = MLAS_GEMM_U8U8_STRIDEN;
-    size_t StrideK = MLAS_GEMM_U8U8_STRIDEK;
+    size_t StrideM = MLAS_GEMM_X8X8_STRIDEM;
+    size_t StrideN = MLAS_GEMM_X8X8_STRIDEN;
+    size_t StrideK = MLAS_GEMM_X8X8_STRIDEK;
 
-    MLAS_UNREFERENCED_PARAMETER(ThreadPool);
+    //
+    // Step through each slice of matrix B along the K dimension.
+    //
 
     size_t CountK;
 
@@ -1238,6 +1344,10 @@ MlasGemm(
             CountK = K - k;
         }
 
+        //
+        // Step through each slice of matrix B along the N dimension.
+        //
+
         size_t CountN;
 
         for (size_t n = 0; n < N; n += CountN) {
@@ -1248,7 +1358,10 @@ MlasGemm(
                 CountN = N - n;
             }
 
-            MlasPlatform.GemmU8U8CopyPackBRoutine(PanelB, B + n + k * ldb, ldb, CountN, CountK, ColumnSumVector, -int16_t(offa));
+            const uint8_t* b = (const uint8_t*)B + n + k * ldb;
+
+            MlasPlatform.GemmU8U8CopyPackBRoutine(PanelB, b, ldb, CountN,
+                CountK, ColumnSumVector, -int16_t(offa));
 
             size_t CountM;
 
@@ -1260,7 +1373,8 @@ MlasGemm(
                     CountM = M - m;
                 }
 
-                MlasPlatform.GemmU8U8CopyPackARoutine(PanelA, A + k + m * lda, lda, CountM, CountK, RowSumVector, -int16_t(offb));
+                MlasPlatform.GemmU8U8CopyPackARoutine(PanelA, A + k + m * lda,
+                    lda, CountM, CountK, RowSumVector, -int16_t(offb));
 
                 int16_t* pa = PanelA;
                 int32_t* c = C + n + m * ldc;
@@ -1274,7 +1388,9 @@ MlasGemm(
 
                 while (RowsRemaining > 0) {
 
-                    RowsHandled = MlasPlatform.GemmU8U8Kernel(pa, PanelB, c, PairCountK, RowsRemaining, CountN, ldc, RowSums, ColumnSumVector, int32_t(CountK) * offa * offb, k == 0);
+                    RowsHandled = MlasPlatform.GemmU8U8Kernel(pa, PanelB, c,
+                        PairCountK, RowsRemaining, CountN, ldc, RowSums,
+                        ColumnSumVector, int32_t(CountK) * offa * offb, k == 0);
 
                     RowsRemaining -= RowsHandled;
                     c += ldc * RowsHandled;
@@ -1286,4 +1402,330 @@ MlasGemm(
     }
 }
 
+void
+MlasGemmX8X8Threaded(
+    void* Context,
+    int32_t ThreadId
+    )
+/*++
+
+Routine Description:
+
+    This routine is invoked from a worker thread to execute a segment of a
+    QGEMM operation.
+
+Arguments:
+
+    Context - Supplies the pointer to the context for the threaded operation.
+
+    ThreadId - Supplies the current index of the threaded operation.
+
+Return Value:
+
+    None.
+
+--*/
+{
+    const auto* WorkBlock = (MLAS_GEMM_X8X8_WORK_BLOCK*)Context;
+
+    const int32_t ThreadCountM = WorkBlock->ThreadCountM;
+    const int32_t ThreadCountN = WorkBlock->ThreadCountN;
+
+    const int32_t ThreadIdM = ThreadId / ThreadCountN;
+    const int32_t ThreadIdN = ThreadId % ThreadCountN;
+
+    //
+    // Partition the operation along the M dimension.
+    //
+
+    size_t M = WorkBlock->M;
+    size_t m;
+    size_t CountM;
+
+    MlasPartitionWork(ThreadIdM, ThreadCountM, M, &m, &CountM);
+
+    //
+    // Partition the operation along the N dimension.
+    //
+
+    size_t N = WorkBlock->N;
+    size_t n;
+    size_t CountN;
+
+    const size_t BlockedN = (N + MLAS_QGEMM_STRIDEN_THREAD_ALIGN - 1) /
+        MLAS_QGEMM_STRIDEN_THREAD_ALIGN;
+
+    MlasPartitionWork(ThreadIdN, ThreadCountN, BlockedN, &n, &CountN);
+
+    n *= MLAS_QGEMM_STRIDEN_THREAD_ALIGN;
+    CountN *= MLAS_QGEMM_STRIDEN_THREAD_ALIGN;
+
+    if (CountN > N - n) {
+        CountN = N - n;
+    }
+
+    //
+    // Dispatch the partitioned operation.
+    //
+
+    const size_t lda = WorkBlock->lda;
+    const size_t ldb = WorkBlock->ldb;
+    const size_t ldc = WorkBlock->ldc;
+
+    const uint8_t* a = WorkBlock->A + m * lda;
+    const uint8_t* b = WorkBlock->B + n;
+    int32_t* c = WorkBlock->C + n + m * ldc;
+
+    WorkBlock->GemmX8X8Operation(CountM, CountN, WorkBlock->K, a, lda,
+        WorkBlock->offa, b, ldb, WorkBlock->offb, c, ldc);
+}
+
+void
+MlasGemmX8X8Schedule(
+    MLAS_GEMM_X8X8_WORK_BLOCK* WorkBlock,
+    MLAS_THREADPOOL* ThreadPool
+    )
+/*++
+
+Routine Description:
+
+    This module schedules the quantized integer matrix/matrix multiply
+    operation (QGEMM) across one or more threads.
+
+Arguments:
+
+    WorkBlock - Supplies the structure containing the GEMM parameters.
+
+    ThreadPool - Supplies the thread pool object to use, else nullptr if the
+        base library threading support should be used.
+
+Return Value:
+
+    None.
+
+--*/
+{
+    const size_t M = WorkBlock->M;
+    const size_t N = WorkBlock->N;
+    const size_t K = WorkBlock->K;
+
+    //
+    // Compute the number of target threads given the complexity of the SGEMM
+    // operation. Small requests should run using the single threaded path.
+    //
+
+    double Complexity = double(M) * double(N) * double(K);
+
+    int32_t TargetThreadCount;
+
+    if (Complexity < double(MLAS_QGEMM_THREAD_COMPLEXITY * MLAS_MAXIMUM_THREAD_COUNT)) {
+        TargetThreadCount = int32_t(Complexity / double(MLAS_QGEMM_THREAD_COMPLEXITY)) + 1;
+    } else {
+        TargetThreadCount = MLAS_MAXIMUM_THREAD_COUNT;
+    }
+
+    int32_t MaximumThreadCount = MlasGetMaximumThreadCount(ThreadPool);
+
+    if (TargetThreadCount >= MaximumThreadCount) {
+        TargetThreadCount = MaximumThreadCount;
+    }
+
+    //
+    // Segment the operation across multiple threads.
+    //
+    // N.B. Currently, the operation is segmented as a 1D partition, which
+    // works okay for operations involving skinny matrices.
+    //
+
+    if (N > M) {
+
+        const size_t BlockedN = (N + MLAS_QGEMM_STRIDEN_THREAD_ALIGN - 1) /
+            MLAS_QGEMM_STRIDEN_THREAD_ALIGN;
+
+        if (size_t(TargetThreadCount) > BlockedN) {
+            TargetThreadCount = int32_t(BlockedN);
+        }
+
+        WorkBlock->ThreadCountM = 1;
+        WorkBlock->ThreadCountN = TargetThreadCount;
+
+    } else {
+
+        if (size_t(TargetThreadCount) > M) {
+            TargetThreadCount = int32_t(M);
+        }
+
+        WorkBlock->ThreadCountM = TargetThreadCount;
+        WorkBlock->ThreadCountN = 1;
+    }
+
+    MlasExecuteThreaded(MlasGemmX8X8Threaded, WorkBlock, TargetThreadCount, ThreadPool);
+}
+
+void
+MLASCALL
+MlasGemm(
+    size_t M,
+    size_t N,
+    size_t K,
+    const uint8_t* A,
+    size_t lda,
+    uint8_t offa,
+    const int8_t* B,
+    size_t ldb,
+    int8_t offb,
+    int32_t* C,
+    size_t ldc,
+    MLAS_THREADPOOL* ThreadPool
+    )
+/*++
+
+Routine Description:
+
+    This module implements the quantized integer matrix/matrix multiply
+    operation (QGEMM).
+
+Arguments:
+
+    M - Supplies the number of rows of matrix A and matrix C.
+
+    N - Supplies the number of columns of matrix B and matrix C.
+
+    K - Supplies the number of columns of matrix A and the number of rows of
+        matrix B.
+
+    A - Supplies the address of matrix A.
+
+    lda - Supplies the first dimension of matrix A.
+
+    offa - Supplies the zero point offset of matrix A.
+
+    B - Supplies the address of matrix B.
+
+    ldb - Supplies the first dimension of matrix B.
+
+    offb - Supplies the zero point offset of matrix B.
+
+    C - Supplies the address of matrix C.
+
+    ldc - Supplies the first dimension of matrix C.
+
+    ThreadPool - Supplies the thread pool object to use, else nullptr if the
+        base library threading support should be used.
+
+Return Value:
+
+    None.
+
+--*/
+{
+    MLAS_GEMM_X8X8_WORK_BLOCK WorkBlock;
+
+    //
+    // Capture the GEMM parameters to the work block.
+    //
+
+    WorkBlock.M = M;
+    WorkBlock.N = N;
+    WorkBlock.K = K;
+    WorkBlock.A = A;
+    WorkBlock.lda = lda;
+    WorkBlock.B = (const uint8_t*)B;
+    WorkBlock.ldb = ldb;
+    WorkBlock.C = C;
+    WorkBlock.ldc = ldc;
+    WorkBlock.offa = int16_t(offa);
+    WorkBlock.offb = int16_t(offb);
+    WorkBlock.GemmX8X8Operation = MlasGemmU8S8Operation;
+
+    //
+    // Schedule the operation across a set of worker threads.
+    //
+
+    MlasGemmX8X8Schedule(&WorkBlock, ThreadPool);
+}
+
+void
+MLASCALL
+MlasGemm(
+    size_t M,
+    size_t N,
+    size_t K,
+    const uint8_t* A,
+    size_t lda,
+    uint8_t offa,
+    const uint8_t* B,
+    size_t ldb,
+    uint8_t offb,
+    int32_t* C,
+    size_t ldc,
+    MLAS_THREADPOOL* ThreadPool
+    )
+/*++
+
+Routine Description:
+
+    This module implements the quantized integer matrix/matrix multiply
+    operation (QGEMM).
+
+Arguments:
+
+    M - Supplies the number of rows of matrix A and matrix C.
+
+    N - Supplies the number of columns of matrix B and matrix C.
+
+    K - Supplies the number of columns of matrix A and the number of rows of
+        matrix B.
+
+    A - Supplies the address of matrix A.
+
+    lda - Supplies the first dimension of matrix A.
+
+    offa - Supplies the zero point offset of matrix A.
+
+    B - Supplies the address of matrix B.
+
+    ldb - Supplies the first dimension of matrix B.
+
+    offb - Supplies the zero point offset of matrix B.
+
+    C - Supplies the address of matrix C.
+
+    ldc - Supplies the first dimension of matrix C.
+
+    ThreadPool - Supplies the thread pool object to use, else nullptr if the
+        base library threading support should be used.
+
+Return Value:
+
+    None.
+
+--*/
+{
+    MLAS_GEMM_X8X8_WORK_BLOCK WorkBlock;
+
+    //
+    // Capture the GEMM parameters to the work block.
+    //
+
+    WorkBlock.M = M;
+    WorkBlock.N = N;
+    WorkBlock.K = K;
+    WorkBlock.A = A;
+    WorkBlock.lda = lda;
+    WorkBlock.B = B;
+    WorkBlock.ldb = ldb;
+    WorkBlock.C = C;
+    WorkBlock.ldc = ldc;
+    WorkBlock.offa = int16_t(offa);
+    WorkBlock.offb = int16_t(offb);
+    WorkBlock.GemmX8X8Operation = MlasGemmU8U8Operation;
+
+    //
+    // Schedule the operation across a set of worker threads.
+    //
+
+    MlasGemmX8X8Schedule(&WorkBlock, ThreadPool);
+}
+
 #endif
diff --git a/onnxruntime/core/mlas/lib/snchwc.cpp b/onnxruntime/core/mlas/lib/snchwc.cpp
index 843f919f44c40..7c3193c2907fb 100644
--- a/onnxruntime/core/mlas/lib/snchwc.cpp
+++ b/onnxruntime/core/mlas/lib/snchwc.cpp
@@ -128,7 +128,7 @@ Routine Description:
 
 Arguments:
 
-    WorkBlock - Supplies the structure that contains the commong convolution
+    WorkBlock - Supplies the structure that contains the common convolution
         and pooling parameters.
 
     Dimensions - Supplies the number of dimensions.
@@ -344,28 +344,6 @@ struct MLAS_NCHWC_NN_ALGORITHM
         OutputCountRightPadX(WorkBlock->OutputCountRightPad[WidthShapeIndex])
     {
     }
-
-    static
-    void
-    PartitionWork(
-        int32_t Index,
-        const MLAS_NCHWC_WORK_BLOCK* WorkBlock,
-        size_t TotalWork,
-        size_t* WorkIndex,
-        size_t* WorkRemaining
-        )
-    {
-        const size_t WorkPerThread = TotalWork / WorkBlock->tids;
-        const size_t WorkPerThreadExtra = TotalWork % WorkBlock->tids;
-
-        if (uint32_t(Index) < WorkPerThreadExtra) {
-            *WorkIndex = (WorkPerThread + 1) * Index;
-            *WorkRemaining = WorkPerThread + 1;
-        } else {
-            *WorkIndex = WorkPerThread * Index + WorkPerThreadExtra;
-            *WorkRemaining = WorkPerThread;
-        }
-    }
 };
 
 constexpr size_t MLAS_NCHWC_NN_ALGORITHM::HeightShapeIndex;
@@ -573,7 +551,7 @@ struct MLAS_NCHWC_GROUPED_CONV_ALGORITHM : MLAS_NCHWC_CONV_ALGORITHM
 
         size_t WorkIndex;
 
-        PartitionWork(Index, WorkBlock, TotalWork, &WorkIndex, &WorkRemaining);
+        MlasPartitionWork(Index, WorkBlock->tids, TotalWork, &WorkIndex, &WorkRemaining);
 
         //
         // Extract the current batch, group, filter cluster, and output line
@@ -1004,7 +982,7 @@ struct MLAS_NCHWC_CONV_DEPTHWISE_ALGORITHM : MLAS_NCHWC_CONV_ALGORITHM
         size_t WorkIndex;
         size_t WorkRemaining;
 
-        PartitionWork(Index, WorkBlock, TotalWork, &WorkIndex, &WorkRemaining);
+        MlasPartitionWork(Index, WorkBlock->tids, TotalWork, &WorkIndex, &WorkRemaining);
 
         //
         // Extract the current batch, group block, and output line from the
@@ -1138,7 +1116,7 @@ struct MLAS_NCHWC_POOL_ALGORITHM : MLAS_NCHWC_NN_ALGORITHM
         size_t WorkIndex;
         size_t WorkRemaining;
 
-        PartitionWork(Index, WorkBlock, TotalWork, &WorkIndex, &WorkRemaining);
+        MlasPartitionWork(Index, WorkBlock->tids, TotalWork, &WorkIndex, &WorkRemaining);
 
         size_t ph = WorkIndex % OutputHeight;
         const size_t BatchChannel = WorkIndex / OutputHeight;
diff --git a/onnxruntime/core/mlas/lib/threading.cpp b/onnxruntime/core/mlas/lib/threading.cpp
index ef30de9499bb2..ab797e431231a 100644
--- a/onnxruntime/core/mlas/lib/threading.cpp
+++ b/onnxruntime/core/mlas/lib/threading.cpp
@@ -46,7 +46,6 @@ MlasExecuteThreaded(
     }
 #endif
 
-
     //
     // Fallback to OpenMP or a serialized implementation.
     //
diff --git a/onnxruntime/core/mlas/lib/x86_64/QgemmU8S8KernelAvx2.S b/onnxruntime/core/mlas/lib/x86_64/QgemmU8S8KernelAvx2.S
index 3d637fb97ec59..6a6edd7e8017b 100644
--- a/onnxruntime/core/mlas/lib/x86_64/QgemmU8S8KernelAvx2.S
+++ b/onnxruntime/core/mlas/lib/x86_64/QgemmU8S8KernelAvx2.S
@@ -130,6 +130,8 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
         vpxor   xmm1,xmm1,xmm1
         vpxor   xmm2,xmm2,xmm2
         vpxor   xmm3,xmm3,xmm3
+        lea     r13,[r10+r10*2]             # compute ldb * 3
+        lea     rax,[r12+r12*2]             # compute output stride * 3
         mov     rdx,rsi
         mov     rcx,rdi
         lea     rsi,[rsi+r10*4]             # advance next matrix A by 4 rows
@@ -139,16 +141,14 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
         jb      .LCopyPackA.ProcessRemainingColumnsM4
 
 .LCopyPackA.ProcessNextColumnLoopM4:
-        lea     rax,[rdx+r10*2]             # compute matrix A plus 2 rows
         vmovdqu ymm4,YMMWORD PTR [rdx]
         vmovdqu ymm5,YMMWORD PTR [rdx+r10]
-        vmovdqu ymm6,YMMWORD PTR [rax]
-        vmovdqu ymm7,YMMWORD PTR [rax+r10]
-        lea     rax,[rcx+r12*2]             # compute matrix D plus 2 rows
+        vmovdqu ymm6,YMMWORD PTR [rdx+r10*2]
+        vmovdqu ymm7,YMMWORD PTR [rdx+r13]
         vmovdqu YMMWORD PTR [rcx],ymm4
         vmovdqu YMMWORD PTR [rcx+r12],ymm5
-        vmovdqu YMMWORD PTR [rax],ymm6
-        vmovdqu YMMWORD PTR [rax+r12],ymm7
+        vmovdqu YMMWORD PTR [rcx+r12*2],ymm6
+        vmovdqu YMMWORD PTR [rcx+rax],ymm7
         vpmaddubsw ymm4,ymm4,ymm9           # horizontal byte+byte=word per row
         vpaddw  ymm0,ymm0,ymm4              # add words to row accumulators
         vpmaddubsw ymm5,ymm5,ymm9
@@ -167,16 +167,14 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
         jz      .LCopyPackA.ReduceRowSumVectorM4
         test    bl,16                       # (CountK & 16) != 0?
         jz      .LCopyPackA.CopyRemainingCountKLessThan16M4
-        lea     rax,[rdx+r10*2]             # compute matrix A plus 2 rows
         vmovdqu xmm4,XMMWORD PTR [rdx]
         vmovdqu xmm5,XMMWORD PTR [rdx+r10]
-        vmovdqu xmm6,XMMWORD PTR [rax]
-        vmovdqu xmm7,XMMWORD PTR [rax+r10]
-        lea     rax,[rcx+r12*2]             # compute matrix D plus 2 rows
+        vmovdqu xmm6,XMMWORD PTR [rdx+r10*2]
+        vmovdqu xmm7,XMMWORD PTR [rdx+r13]
         vmovdqu XMMWORD PTR [rcx],xmm4
         vmovdqu XMMWORD PTR [rcx+r12],xmm5
-        vmovdqu XMMWORD PTR [rax],xmm6
-        vmovdqu XMMWORD PTR [rax+r12],xmm7
+        vmovdqu XMMWORD PTR [rcx+r12*2],xmm6
+        vmovdqu XMMWORD PTR [rcx+rax],xmm7
         vpmaddubsw xmm4,xmm4,xmm9           # horizontal byte+byte=word per row
         vpaddw  ymm0,ymm0,ymm4              # add words to row accumulators
         vpmaddubsw xmm5,xmm5,xmm9
@@ -198,14 +196,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
         lea     rbp,.LGemmU8S8CopyPackAFrame_PaddedMatrixAData[rsp]
         test    bl,8                        # (CountK & 8) != 0?
         jz      .LCopyPackA.CopyRemainingCountKLessThan8M4
-        lea     r13,[rdx+r10*2]             # compute matrix A plus 2 rows
         mov     rax,QWORD PTR [rdx]
         mov     QWORD PTR [rbp],rax
         mov     rax,QWORD PTR [rdx+r10]
         mov     QWORD PTR [rbp+16],rax
-        mov     rax,QWORD PTR [r13]
+        mov     rax,QWORD PTR [rdx+r10*2]
         mov     QWORD PTR [rbp+32],rax
-        mov     rax,QWORD PTR [r13+r10]
+        mov     rax,QWORD PTR [rdx+r13]
         mov     QWORD PTR [rbp+48],rax
         add     rdx,8
         add     rbp,8                       # advance padded buffer destination
@@ -213,14 +210,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
 .LCopyPackA.CopyRemainingCountKLessThan8M4:
         test    bl,4                        # (CountK & 4) != 0?
         jz      .LCopyPackA.CopyRemainingCountKLessThan4M4
-        lea     r13,[rdx+r10*2]             # compute matrix A plus 2 rows
         mov     eax,DWORD PTR [rdx]
         mov     DWORD PTR [rbp],eax
         mov     eax,DWORD PTR [rdx+r10]
         mov     DWORD PTR [rbp+16],eax
-        mov     eax,DWORD PTR [r13]
+        mov     eax,DWORD PTR [rdx+r10*2]
         mov     DWORD PTR [rbp+32],eax
-        mov     eax,DWORD PTR [r13+r10]
+        mov     eax,DWORD PTR [rdx+r13]
         mov     DWORD PTR [rbp+48],eax
         add     rdx,4
         add     rbp,4                       # advance padded buffer destination
@@ -228,14 +224,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
 .LCopyPackA.CopyRemainingCountKLessThan4M4:
         test    bl,2                        # (CountK & 2) != 0?
         jz      .LCopyPackA.CopyRemainingCountKLessThan2M4
-        lea     r13,[rdx+r10*2]             # compute matrix A plus 2 rows
         movzx   eax,WORD PTR [rdx]
         mov     WORD PTR [rbp],ax
         movzx   eax,WORD PTR [rdx+r10]
         mov     WORD PTR [rbp+16],ax
-        movzx   eax,WORD PTR [r13]
+        movzx   eax,WORD PTR [rdx+r10*2]
         mov     WORD PTR [rbp+32],ax
-        movzx   eax,WORD PTR [r13+r10]
+        movzx   eax,WORD PTR [rdx+r13]
         mov     WORD PTR [rbp+48],ax
         add     rdx,2
         add     rbp,2                       # advance padded buffer destination
@@ -243,14 +238,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackAAvx2):
 .LCopyPackA.CopyRemainingCountKLessThan2M4:
         test    bl,1                        # (CountK & 1) != 0?
         jz      .LCopyPackA.ProcessPaddedMatrixADataM4
-        lea     r13,[rdx+r10*2]             # compute matrix A plus 2 rows
         movzx   eax,BYTE PTR [rdx]
         mov     BYTE PTR [rbp],al
         movzx   eax,BYTE PTR [rdx+r10]
         mov     BYTE PTR [rbp+16],al
-        movzx   eax,BYTE PTR [r13]
+        movzx   eax,BYTE PTR [rdx+r10*2]
         mov     BYTE PTR [rbp+32],al
-        movzx   eax,BYTE PTR [r13+r10]
+        movzx   eax,BYTE PTR [rdx+r13]
         mov     BYTE PTR [rbp+48],al
 
 //
@@ -446,6 +440,7 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackBAvx2):
         push    rbx
 
         mov     r10,rdx
+        lea     r11,[r10+r10*2]             # compute ldb * 3
         vpbroadcastw ymm7,WORD PTR .LGemmU8S8CopyPackBFrame_offa[rsp]
         vpcmpeqw ymm8,ymm8,ymm8             # generate word vector [0xFFFF]
         vpsrlw  ymm8,ymm8,15                # generate word vector [0x0001]
@@ -469,11 +464,10 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackBAvx2):
         jb      .LCopyPackB.ProcessRemainingRowsN16
 
 .LCopyPackB.ProcessNextRowLoopN16:
-        lea     rax,[rdx+r10*2]             # compute matrix B plus 2 rows
         vmovdqu xmm2,XMMWORD PTR [rdx]      # load 4 rows
         vmovdqu xmm3,XMMWORD PTR [rdx+r10]
-        vmovdqu xmm4,XMMWORD PTR [rax]
-        vmovdqu xmm5,XMMWORD PTR [rax+r10]
+        vmovdqu xmm4,XMMWORD PTR [rdx+r10*2]
+        vmovdqu xmm5,XMMWORD PTR [rdx+r11]
         lea     rdx,[rdx+r10*4]             # advance matrix B by 4 rows
 
 .LCopyPackB.InterleaveRowDataN16:
@@ -558,14 +552,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackBAvx2):
         lea     rbp,.LGemmU8S8CopyPackBFrame_PaddedMatrixBData[rsp]
         test    cl,8                        # (CountN & 8) != 0?
         jz      .LCopyPackB.CopyRemainingCountNLessThan8K4
-        lea     r11,[rdx+r10*2]             # compute matrix B plus 2 rows
         mov     rax,QWORD PTR [rdx]
         mov     QWORD PTR [rbp],rax
         mov     rax,QWORD PTR [rdx+r10]
         mov     QWORD PTR [rbp+16],rax
-        mov     rax,QWORD PTR [r11]
+        mov     rax,QWORD PTR [rdx+r10*2]
         mov     QWORD PTR [rbp+32],rax
-        mov     rax,QWORD PTR [r11+r10]
+        mov     rax,QWORD PTR [rdx+r11]
         mov     QWORD PTR [rbp+48],rax
         add     rdx,8                       # advance matrix B
         add     rbp,8                       # advance padded buffer destination
@@ -573,14 +566,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackBAvx2):
 .LCopyPackB.CopyRemainingCountNLessThan8K4:
         test    cl,4                        # (CountN & 4) != 0?
         jz      .LCopyPackB.CopyRemainingCountNLessThan4K4
-        lea     r11,[rdx+r10*2]             # compute matrix B plus 2 rows
         mov     eax,DWORD PTR [rdx]
         mov     DWORD PTR [rbp],eax
         mov     eax,DWORD PTR [rdx+r10]
         mov     DWORD PTR [rbp+16],eax
-        mov     eax,DWORD PTR [r11]
+        mov     eax,DWORD PTR [rdx+r10*2]
         mov     DWORD PTR [rbp+32],eax
-        mov     eax,DWORD PTR [r11+r10]
+        mov     eax,DWORD PTR [rdx+r11]
         mov     DWORD PTR [rbp+48],eax
         add     rdx,4                       # advance matrix B
         add     rbp,4                       # advance padded buffer destination
@@ -588,14 +580,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackBAvx2):
 .LCopyPackB.CopyRemainingCountNLessThan4K4:
         test    cl,2                        # (CountN & 2) != 0?
         jz      .LCopyPackB.CopyRemainingCountNLessThan2K4
-        lea     r11,[rdx+r10*2]             # compute matrix B plus 2 rows
         movzx   eax,WORD PTR [rdx]
         mov     WORD PTR [rbp],ax
         movzx   eax,WORD PTR [rdx+r10]
         mov     WORD PTR [rbp+16],ax
-        movzx   eax,WORD PTR [r11]
+        movzx   eax,WORD PTR [rdx+r10*2]
         mov     WORD PTR [rbp+32],ax
-        movzx   eax,WORD PTR [r11+r10]
+        movzx   eax,WORD PTR [rdx+r11]
         mov     WORD PTR [rbp+48],ax
         add     rdx,2                       # advance matrix B
         add     rbp,2                       # advance padded buffer destination
@@ -603,14 +594,13 @@ C_UNDERSCORE(MlasGemmU8S8CopyPackBAvx2):
 .LCopyPackB.CopyRemainingCountNLessThan2K4:
         test    cl,1                        # (CountN & 1) != 0?
         jz      .LCopyPackB.ProcessPaddedMatrixBData
-        lea     r11,[rdx+r10*2]             # compute matrix B plus 2 rows
         movzx   eax,BYTE PTR [rdx]
         mov     BYTE PTR [rbp],al
         movzx   eax,BYTE PTR [rdx+r10]
         mov     BYTE PTR [rbp+16],al
-        movzx   eax,BYTE PTR [r11]
+        movzx   eax,BYTE PTR [rdx+r10*2]
         mov     BYTE PTR [rbp+32],al
-        movzx   eax,BYTE PTR [r11+r10]
+        movzx   eax,BYTE PTR [rdx+r11]
         mov     BYTE PTR [rbp+48],al
 
 .LCopyPackB.ProcessPaddedMatrixBData:
diff --git a/onnxruntime/core/providers/cpu/math/gemm.h b/onnxruntime/core/providers/cpu/math/gemm.h
index a309ad144a01e..227bf46c02d1e 100644
--- a/onnxruntime/core/providers/cpu/math/gemm.h
+++ b/onnxruntime/core/providers/cpu/math/gemm.h
@@ -8,7 +8,6 @@
 #include "core/util/math.h"
 #include "core/util/math_cpuonly.h"
 #include "gemm_helper.h"
-#include "core/framework/op_kernel_context_internal.h"
 
 namespace onnxruntime {
 
@@ -28,7 +27,7 @@ class Gemm : public OpKernel {
   }
 
   Status Compute(OpKernelContext* context) const override {
-    concurrency::ThreadPool* tp = context->GetOperatorThreadPool();
+    concurrency::ThreadPool* thread_pool = context->GetOperatorThreadPool();
 
     const auto* X = context->Input<Tensor>(0);
     const auto* W = context->Input<Tensor>(1);
@@ -82,7 +81,7 @@ class Gemm : public OpKernel {
         // but passing 0 for beta is cheaper and it will ignore any junk in the output buffer
         B != nullptr ? beta_ : 0,
         y_data,
-        tp);
+        thread_pool);
 
     FuseActivation<T>(activation_, y_data, M * N, leaky_relu_alpha_);
 
diff --git a/onnxruntime/core/providers/cpu/math/matmul.cc b/onnxruntime/core/providers/cpu/math/matmul.cc
index cff1b1066a26f..5e4c5fb2a55ef 100644
--- a/onnxruntime/core/providers/cpu/math/matmul.cc
+++ b/onnxruntime/core/providers/cpu/math/matmul.cc
@@ -1,8 +1,7 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
-#include "core/framework/op_kernel_context_internal.h"
-#include "core/providers/cpu/math/matmul.h"
 
+#include "core/providers/cpu/math/matmul.h"
 #include "core/util/math.h"
 #include "core/util/math_cpuonly.h"
 #include "matmul_helper.h"
diff --git a/onnxruntime/core/providers/cpu/math/matmul_integer.cc b/onnxruntime/core/providers/cpu/math/matmul_integer.cc
index b76470e098928..26bfaaa013970 100644
--- a/onnxruntime/core/providers/cpu/math/matmul_integer.cc
+++ b/onnxruntime/core/providers/cpu/math/matmul_integer.cc
@@ -2,7 +2,6 @@
 // Licensed under the MIT License.
 
 #include "core/framework/data_types_internal.h"
-#include "core/framework/op_kernel_context_internal.h"
 #include "core/providers/cpu/math/matmul_integer.h"
 #include "core/providers/cpu/math/matmul_helper.h"
 #include "core/util/qmath.h"
diff --git a/onnxruntime/core/providers/cpu/nn/conv.cc b/onnxruntime/core/providers/cpu/nn/conv.cc
index f7023a98d3209..e213db54fc426 100644
--- a/onnxruntime/core/providers/cpu/nn/conv.cc
+++ b/onnxruntime/core/providers/cpu/nn/conv.cc
@@ -78,7 +78,7 @@ Status Conv<T>::Compute(OpKernelContext* context) const {
                           output_shape.GetDims().end());
 
   const size_t kernel_rank = kernel_shape.size();
-  concurrency::ThreadPool* tp = context->GetOperatorThreadPool();
+  concurrency::ThreadPool* thread_pool = context->GetOperatorThreadPool();
 
   for (int image_id = 0; image_id < N; ++image_id) {
     for (int group_id = 0; group_id < conv_attrs_.group; ++group_id) {
@@ -125,7 +125,7 @@ Status Conv<T>::Compute(OpKernelContext* context) const {
           col_buffer_data,
           0,
           Ydata + group_id * Y_offset,
-          tp);
+          thread_pool);
     }
 
     if (B != nullptr) {
@@ -186,7 +186,7 @@ Status Conv<float>::Compute(OpKernelContext* context) const {
   auto* Ydata = Y->template MutableData<float>();
 
   const size_t kernel_rank = kernel_shape.size();
-  concurrency::ThreadPool* tp = context->GetOperatorThreadPool();
+  concurrency::ThreadPool* thread_pool = context->GetOperatorThreadPool();
 
   if (kernel_rank == 2 || kernel_rank == 3) {
     MLAS_CONV_PARAMETERS Parameters;
@@ -205,7 +205,7 @@ Status Conv<float>::Compute(OpKernelContext* context) const {
                     static_cast<size_t>(M / conv_attrs_.group),
                     &activation_,
                     &WorkingBufferSize,
-                    tp);
+                    thread_pool);
 
     auto working_data = WorkingBufferSize > 0 ? alloc->Alloc(sizeof(float) * WorkingBufferSize) : nullptr;
     BufferUniquePtr working_buffer(working_data, BufferDeleter(alloc));
@@ -216,7 +216,7 @@ Status Conv<float>::Compute(OpKernelContext* context) const {
              Bdata,
              static_cast<float*>(working_buffer.get()),
              Ydata,
-             tp);
+             thread_pool);
   } else {
     const int64_t input_image_size = input_shape.Size();
     const int64_t output_image_size = output_shape.Size();
@@ -262,7 +262,7 @@ Status Conv<float>::Compute(OpKernelContext* context) const {
             col_buffer_data,
             0,
             Ydata + group_id * Y_offset,
-            tp);
+            thread_pool);
       }
 
       MlasActivation(&activation_, Ydata, Bdata, M, output_image_size, output_image_size);
diff --git a/onnxruntime/core/providers/cpu/nn/conv_integer.cc b/onnxruntime/core/providers/cpu/nn/conv_integer.cc
index 7e5951d4ec699..53ce6e1070f6a 100644
--- a/onnxruntime/core/providers/cpu/nn/conv_integer.cc
+++ b/onnxruntime/core/providers/cpu/nn/conv_integer.cc
@@ -90,6 +90,7 @@ Status ConvInteger::Compute(OpKernelContext* context) const {
                           output_shape.GetDims().end());
 
   const size_t kernel_rank = kernel_shape.size();
+  concurrency::ThreadPool* thread_pool = context->GetOperatorThreadPool();
 
   for (int image_id = 0; image_id < N; ++image_id) {
     for (int group_id = 0; group_id < conv_attrs_.group; ++group_id) {
@@ -139,7 +140,7 @@ Status ConvInteger::Compute(OpKernelContext* context) const {
                     input_offset,
                     Ydata + group_id * Y_offset,
                     static_cast<int>(output_image_size),
-                    nullptr);
+                    thread_pool);
     }
 
     Xdata += X_offset * conv_attrs_.group;
diff --git a/onnxruntime/core/providers/cpu/nn/conv_transpose.cc b/onnxruntime/core/providers/cpu/nn/conv_transpose.cc
index e025c12979e36..ab763b8d9f788 100644
--- a/onnxruntime/core/providers/cpu/nn/conv_transpose.cc
+++ b/onnxruntime/core/providers/cpu/nn/conv_transpose.cc
@@ -16,7 +16,6 @@
 /* Modifications Copyright (c) Microsoft. */
 
 #include "core/providers/cpu/nn/conv_transpose.h"
-#include "core/framework/op_kernel_context_internal.h"
 
 #include "core/util/math.h"
 #include "core/util/math_cpuonly.h"
@@ -42,7 +41,7 @@ Status ConvTranspose<T>::Compute(OpKernelContext* context) const {
 
 template <typename T>
 Status ConvTranspose<T>::DoConvTranspose(OpKernelContext* context, bool dynamic_padding) const {
-  concurrency::ThreadPool* tp = context->GetOperatorThreadPool();
+  concurrency::ThreadPool* thread_pool = context->GetOperatorThreadPool();
 
   size_t num_inputs = OpKernel::Node().InputDefs().size();
   ConvTransposeAttributes::Prepare p;
@@ -87,7 +86,7 @@ Status ConvTranspose<T>::DoConvTranspose(OpKernelContext* context, bool dynamic_
             Xdata + group_id * X_offset,
             0,
             col_buffer_data,
-            tp);
+            thread_pool);
 
         // Col2im
         math::Col2im<T, CPUMathUtil, StorageOrder::NCHW>(
@@ -136,7 +135,7 @@ Status ConvTranspose<T>::DoConvTranspose(OpKernelContext* context, bool dynamic_
             Xdata + group_id * X_offset,
             0,
             col_buffer_data,
-            tp);
+            thread_pool);
 
         // Col2im
         math::Col2imNd<T, CPUMathUtil, StorageOrder::NCHW>(
@@ -155,7 +154,7 @@ Status ConvTranspose<T>::DoConvTranspose(OpKernelContext* context, bool dynamic_
       }
 
       if (p.B != nullptr) {
-        
+
         auto Ymatrix = EigenMatrixMap<T>(Ydata, output_size, p.num_output_channels);
         auto Bvec = ConstEigenVectorMap<T>(p.B->template Data<T>(), p.num_output_channels);
         Ymatrix.rowwise() += Bvec.transpose();
diff --git a/onnxruntime/core/providers/cpu/nn/pool.cc b/onnxruntime/core/providers/cpu/nn/pool.cc
index ec7a5904c754d..8bf630eb056a9 100644
--- a/onnxruntime/core/providers/cpu/nn/pool.cc
+++ b/onnxruntime/core/providers/cpu/nn/pool.cc
@@ -1,7 +1,6 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
-#include "core/framework/op_kernel_context_internal.h"
 #include "core/providers/cpu/nn/pool.h"
 
 using namespace ::onnxruntime::common;

From 45c58356d7f5aa50d938b843ec8b3b545c69b3d1 Mon Sep 17 00:00:00 2001
From: Tiago Koji Castro Shibata <ticastro@microsoft.com>
Date: Thu, 16 Jan 2020 00:19:46 -0800
Subject: [PATCH 69/75] Fix test warnings and delayload linking (#2843)

---
 cmake/winml.cmake                              |  6 +++---
 cmake/winml_unittests.cmake                    |  4 ++--
 winml/test/api/LearningModelSessionAPITest.cpp |  2 +-
 winml/test/common/googleTestMacros.h           |  2 +-
 winml/test/common/taefTestMacros.h             | 14 ++++++++------
 winml/test/common/test.h                       | 10 +++++++++-
 6 files changed, 24 insertions(+), 14 deletions(-)

diff --git a/cmake/winml.cmake b/cmake/winml.cmake
index f5171158f230e..f9c02ddf0e9e5 100644
--- a/cmake/winml.cmake
+++ b/cmake/winml.cmake
@@ -442,15 +442,15 @@ endif(onnxruntime_USE_DML)
 
 # The default libraries to link with in Windows are kernel32.lib;user32.lib;gdi32.lib;winspool.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;comdlg32.lib;advapi32.lib
 # Remove them and use the onecore umbrella library instead
-set(CMAKE_C_STANDARD_LIBRARIES "onecoreuap_apiset.lib")
-set(CMAKE_CXX_STANDARD_LIBRARIES "onecoreuap_apiset.lib")
 foreach(default_lib kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdgl32.lib advapi32.lib)
   set(removed_libs "${removed_libs} /NODEFAULTLIB:${default_lib}")
 endforeach()
+set(CMAKE_C_STANDARD_LIBRARIES "${removed_libs} onecoreuap.lib")
+set(CMAKE_CXX_STANDARD_LIBRARIES "${removed_libs} onecoreuap.lib")
 set_target_properties(winml_dll
     PROPERTIES
     LINK_FLAGS
-    "/DEF:${WINML_DIR}/windows.ai.machinelearning.def ${os_component_link_flags} /DELAYLOAD:d3d12.dll /DELAYLOAD:d3d11.dll /DELAYLOAD:dxgi.dll ${delayload_dml} ${removed_libs}")
+    "/DEF:${WINML_DIR}/windows.ai.machinelearning.def ${os_component_link_flags} /DELAYLOAD:d3d12.dll /DELAYLOAD:d3d11.dll /DELAYLOAD:dxgi.dll ${delayload_dml}")
 
 
 set_target_properties(winml_dll
diff --git a/cmake/winml_unittests.cmake b/cmake/winml_unittests.cmake
index bd91611a6d17c..0596c711dca1a 100644
--- a/cmake/winml_unittests.cmake
+++ b/cmake/winml_unittests.cmake
@@ -44,7 +44,7 @@ function(add_winml_test)
   if (_UT_DEPENDS)
     add_dependencies(${_UT_TARGET} ${_UT_DEPENDS})
   endif()
-  target_link_libraries(${_UT_TARGET} PRIVATE ${_UT_LIBS} gtest windowsapp winml_lib_image ${onnxruntime_EXTERNAL_LIBRARIES} winml_lib_telemetry winml_lib_api onnxruntime)
+  target_link_libraries(${_UT_TARGET} PRIVATE ${_UT_LIBS} gtest winml_lib_image ${onnxruntime_EXTERNAL_LIBRARIES} winml_lib_telemetry winml_lib_api onnxruntime)
 
   add_test(NAME ${_UT_TARGET}
     COMMAND ${_UT_TARGET}
@@ -81,7 +81,7 @@ endif()
 add_winml_test(
   TARGET winml_test_scenario
   SOURCES ${winml_test_scenario_src}
-  LIBS winml_test_common ${winml_test_scenario_libs}
+  LIBS winml_test_common delayimp.lib ${winml_test_scenario_libs}
 )
 target_precompiled_header(winml_test_scenario testPch.h)
 target_compile_definitions(winml_test_scenario PRIVATE BUILD_GOOGLE_TEST)
diff --git a/winml/test/api/LearningModelSessionAPITest.cpp b/winml/test/api/LearningModelSessionAPITest.cpp
index 6ee0b3d57f751..908d13532e44d 100644
--- a/winml/test/api/LearningModelSessionAPITest.cpp
+++ b/winml/test/api/LearningModelSessionAPITest.cpp
@@ -417,4 +417,4 @@ const LearningModelSesssionAPITestApi& getapi() {
     CloseSession,
   };
  return api;
-}
\ No newline at end of file
+}
diff --git a/winml/test/common/googleTestMacros.h b/winml/test/common/googleTestMacros.h
index f0507435b95c8..b5199df0bb668 100644
--- a/winml/test/common/googleTestMacros.h
+++ b/winml/test/common/googleTestMacros.h
@@ -66,7 +66,7 @@
 
 #ifndef USE_DML
 #define GPUTEST \
-  WINML_SKIP_TEST("GPU tests disabled because this is a WinML only build (no DML)")
+  WINML_SUPRESS_UNREACHABLE_BELOW(WINML_SKIP_TEST("GPU tests disabled because this is a WinML only build (no DML)"))
 #else
 #define GPUTEST                                                                         \
   if (auto noGpuTests = RuntimeParameters::Parameters.find("noGPUtests");               \
diff --git a/winml/test/common/taefTestMacros.h b/winml/test/common/taefTestMacros.h
index 06d9048d7c6ff..083baa531a99e 100644
--- a/winml/test/common/taefTestMacros.h
+++ b/winml/test/common/taefTestMacros.h
@@ -23,10 +23,12 @@ using namespace WEX::TestExecution;
     getapi().test_name();                 \
   }
 
-#define WINML_SKIP_TEST(message)                                                               \
-  Log::Result(TestResults::Skipped,                                                            \
-              std::wstring_convert<std::codecvt_utf8<wchar_t>>().from_bytes(message).c_str()); \
-  return;
+#define WINML_SKIP_TEST(message)                                                                 \
+  do {                                                                                           \
+    Log::Result(TestResults::Skipped,                                                            \
+                std::wstring_convert<std::codecvt_utf8<wchar_t>>().from_bytes(message).c_str()); \
+    return;                                                                                      \
+  } while (0)
 
 #define WINML_EXPECT_NO_THROW(statement) VERIFY_NO_THROW(statement)
 #define WINML_EXPECT_TRUE(statement) VERIFY_IS_TRUE(statement)
@@ -43,7 +45,7 @@ using namespace WEX::TestExecution;
 
 #ifndef USE_DML
 #define GPUTEST \
-  WINML_SKIP_TEST("GPU tests disabled because this is a WinML only build (no DML)")
+  WINML_SUPRESS_UNREACHABLE_BELOW(WINML_SKIP_TEST("GPU tests disabled because this is a WinML only build (no DML)"))
 #else
 #define GPUTEST                                                                             \
   bool noGPUTests;                                                                          \
@@ -58,4 +60,4 @@ using namespace WEX::TestExecution;
   if (SUCCEEDED(RuntimeParameters::TryGetValue(L"EdgeCore", edgeCoreRun)) && edgeCoreRun) { \
     WINML_SKIP_TEST("This test is disabled by the EdgeCore runtime parameter.");            \
     return;                                                                                 \
-  }
\ No newline at end of file
+  }
diff --git a/winml/test/common/test.h b/winml/test/common/test.h
index 31739f672503f..037f59e932484 100644
--- a/winml/test/common/test.h
+++ b/winml/test/common/test.h
@@ -1,10 +1,18 @@
+#pragma once
+
 using VoidTest = void (*)();
 using SetupTest = VoidTest;
 
+constexpr bool alwaysTrue() {
+    return true;
+}
+#define WINML_SUPRESS_UNREACHABLE_BELOW(statement)    \
+    if (alwaysTrue()) { statement; }
+
 #ifdef BUILD_GOOGLE_TEST
 #include "googleTestMacros.h"
 #else
 #ifdef BUILD_TAEF_TEST
 #include "taefTestMacros.h"
 #endif
-#endif
\ No newline at end of file
+#endif

From 650bdf95339fe10e87f2db160a71c3fb88279c58 Mon Sep 17 00:00:00 2001
From: Ryan Lai <rylai@microsoft.com>
Date: Thu, 16 Jan 2020 11:05:49 -0800
Subject: [PATCH 70/75] Ortmemoryinfo struct changed

---
 winml/adapter/WinMLAdapter.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/winml/adapter/WinMLAdapter.cpp b/winml/adapter/WinMLAdapter.cpp
index 56e7cc043e58e..eef56c60799a8 100644
--- a/winml/adapter/WinMLAdapter.cpp
+++ b/winml/adapter/WinMLAdapter.cpp
@@ -594,7 +594,7 @@ class WinMLAdapter : public Microsoft::WRL::RuntimeClass<
     auto allocator = provider->GetAllocator(0, ::OrtMemType::OrtMemTypeDefault);
 
     const auto& info = allocator->Info();
-    *memory_info = new OrtMemoryInfo(info.name, info.type, info.device, info.id, info.mem_type);
+    *memory_info = new OrtMemoryInfo(info.name, info.alloc_type, info.device, info.id, info.mem_type);
     if (*memory_info == nullptr) {
       return E_OUTOFMEMORY;
     }
@@ -605,7 +605,7 @@ class WinMLAdapter : public Microsoft::WRL::RuntimeClass<
   HRESULT STDMETHODCALLTYPE GetValueMemoryInfo(const OrtValue* ort_value, OrtMemoryInfo** memory_info) override try {
     const auto& tensor = ort_value->Get<onnxruntime::Tensor>();
     auto info = tensor.Location();
-    *memory_info = new OrtMemoryInfo(info.name, info.type, info.device, info.id, info.mem_type);
+    *memory_info = new OrtMemoryInfo(info.name, info.alloc_type, info.device, info.id, info.mem_type);
     if (*memory_info == nullptr) {
       return E_OUTOFMEMORY;
     }

From 5d3f05d8563945e21e2b28931e6795d68f4cb875 Mon Sep 17 00:00:00 2001
From: Ori Levari <ori.levari@microsoft.com>
Date: Thu, 16 Jan 2020 11:48:44 -0800
Subject: [PATCH 71/75] mark the camera scenario test as edgecore because it
 uses d3d11 (#2852)

---
 winml/test/scenario/cppwinrt/scenariotestscppwinrt.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h
index 9e49cf1d3752d..f3d335eb3ba45 100644
--- a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h
+++ b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.h
@@ -61,7 +61,6 @@ WINML_TEST(ScenarioCppWinrtGpuTest, Scenario6BindWithProperties)
 WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleDefaultDirectX)
 WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMinPower)
 WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMaxPerf)
-WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleMyCameraDevice)
 WINML_TEST(ScenarioCppWinrtGpuTest, Scenario8SetDeviceSampleCustomCommandQueue)
 WINML_TEST(ScenarioCppWinrtGpuTest, DISABLED_Scenario9LoadBindEvalInputTensorGPU)
 WINML_TEST(ScenarioCppWinrtGpuTest, Scenario13SingleModelOnCPUandGPU)
@@ -80,6 +79,7 @@ WINML_TEST(ScenarioCppWinrtGpuTest, DeviceLostRecovery)
 WINML_TEST_CLASS_END()
 
 WINML_TEST_CLASS_BEGIN_WITH_SETUP(ScenarioCppWinrtGpuSkipEdgeCoreTest, ScenarioCppWinrtGpuSkipEdgeCoreTestSetup)
+WINML_TEST(ScenarioCppWinrtGpuSkipEdgeCoreTest, Scenario8SetDeviceSampleMyCameraDevice)
 WINML_TEST(ScenarioCppWinrtGpuSkipEdgeCoreTest, Scenario8SetDeviceSampleD3D11Device )
 WINML_TEST(ScenarioCppWinrtGpuSkipEdgeCoreTest, D2DInterop)
 WINML_TEST_CLASS_END()
\ No newline at end of file

From 6dbfc5a33889c46de7ee80e0e417659d8d32c87d Mon Sep 17 00:00:00 2001
From: Ori Levari <ori.levari@microsoft.com>
Date: Thu, 16 Jan 2020 14:14:54 -0800
Subject: [PATCH 72/75] User/orilevari/pipeline fi breaks (#2853)

* remove conflicting artifact names. Decided to stop using drop-nuget-cuda since this may have implications on other dependent pipelines.

* change job name in gpu.yml back to Windows_CI_GPU_CUDA_Dev
---
 tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
index f560a632c94f8..fd7e38df2a600 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
@@ -12,7 +12,7 @@ jobs:
   parameters:
     AgentPool : 'Win-GPU-2019'
     ArtifactName: 'drop-nuget'    
-    JobName: 'Windows_CI_GPU_Dev'
+    JobName: 'Windows_CI_GPU_CUDA_Dev'
     BuildCommand: --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --use_featurizers --enable_onnx_tests --use_telemetry --cmake_generator "Visual Studio 16 2019" --msvc_toolset 14.16 --use_cuda --cuda_version=10.0 --cuda_home="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda"
     BuildArch: 'x64'
     msbuildArchitecture: 'amd64'
@@ -23,7 +23,6 @@ jobs:
     DoCompliance: 'false'
     DoEsrp: ${{ parameters.DoEsrp }}    
     CudaVersion: '10.0'
-    ArtifactName: drop-nuget-cuda
     OrtPackageId: 'Microsoft.ML.OnnxRuntime.Gpu'
     NuPackScript: |
      msbuild $(Build.SourcesDirectory)\csharp\OnnxRuntime.CSharp.proj /p:Configuration=RelWithDebInfo /t:CreatePackage /p:OrtPackageId=Microsoft.ML.OnnxRuntime.Gpu

From 6dee13d5affc0fee3d92b9bac8d7b3dab30fc100 Mon Sep 17 00:00:00 2001
From: Tiago Koji Castro Shibata <ticastro@microsoft.com>
Date: Fri, 17 Jan 2020 14:47:46 -0800
Subject: [PATCH 73/75] Remove internal libs from tests (#2864)

---
 cmake/winml.cmake                             |  30 ++-
 cmake/winml_unittests.cmake                   |   3 +-
 winml/lib/Api.Image/D3DDeviceCache.cpp        |   9 +-
 winml/lib/Api.Image/DeviceHelpers.cpp         | 209 +----------------
 winml/lib/Api.Image/inc/DeviceHelpers.h       |  19 +-
 winml/lib/Common/CommonDeviceHelpers.cpp      | 215 ++++++++++++++++++
 winml/lib/Common/inc/CommonDeviceHelpers.h    |  50 ++++
 .../test/api/LearningModelBindingAPITest.cpp  |   3 +-
 .../test/api/LearningModelSessionAPITest.cpp  |   8 +-
 winml/test/common/std.h                       |   3 +-
 .../cppwinrt/scenariotestscppwinrt.cpp        |  17 +-
 11 files changed, 318 insertions(+), 248 deletions(-)
 create mode 100644 winml/lib/Common/CommonDeviceHelpers.cpp
 create mode 100644 winml/lib/Common/inc/CommonDeviceHelpers.h

diff --git a/cmake/winml.cmake b/cmake/winml.cmake
index f9c02ddf0e9e5..2958a16681793 100644
--- a/cmake/winml.cmake
+++ b/cmake/winml.cmake
@@ -258,7 +258,7 @@ add_dependencies(winml_lib_image winml_api_native)
 add_dependencies(winml_lib_image winml_api_native_internal)
 
 # Link libraries
-target_link_libraries(winml_lib_image PRIVATE wil)
+target_link_libraries(winml_lib_image PRIVATE wil winml_lib_common)
 if (onnxruntime_USE_DML)
   target_add_dml(winml_lib_image)
 endif(onnxruntime_USE_DML)
@@ -360,11 +360,37 @@ add_dependencies(winml_lib_api winml_api_native)
 add_dependencies(winml_lib_api winml_api_native_internal)
 
 # Link libraries
-target_link_libraries(winml_lib_api PRIVATE wil)
+target_link_libraries(winml_lib_api PRIVATE wil winml_lib_telemetry)
 if (onnxruntime_USE_DML)
   target_add_dml(winml_lib_api)
 endif(onnxruntime_USE_DML)
 
+###########################
+# Add winml_lib_common
+###########################
+
+add_library(winml_lib_common STATIC
+  ${winml_lib_common_dir}/CommonDeviceHelpers.cpp
+)
+
+set_target_properties(winml_lib_common PROPERTIES CXX_STANDARD 17)
+set_target_properties(winml_lib_common PROPERTIES CXX_STANDARD_REQUIRED ON)
+target_compile_options(winml_lib_common PRIVATE /GR- /await /bigobj /wd4238)
+target_link_libraries(winml_lib_common PRIVATE wil)
+target_include_directories(winml_lib_common PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api)
+target_compile_definitions(winml_lib_common PRIVATE
+  ONNX_NAMESPACE=onnx
+  ONNX_ML
+  LOTUS_LOG_THRESHOLD=2
+  LOTUS_ENABLE_STDERR_LOGGING
+  PLATFORM_WINDOWS
+  _SCL_SECURE_NO_WARNINGS)
+add_dependencies(winml_lib_common winml_sdk_cppwinrt)
+
+target_include_directories(winml_lib_common PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api)                   # windows machine learning generated component headers
+target_include_directories(winml_lib_common PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/winml_api/comp_generated)    # windows machine learning generated component headers
+target_include_directories(winml_lib_common PRIVATE ${winml_lib_api_dir})
+target_include_directories(winml_lib_common PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 
 ###########################
 # Add winml_dll
diff --git a/cmake/winml_unittests.cmake b/cmake/winml_unittests.cmake
index 0596c711dca1a..29ee21ce742c4 100644
--- a/cmake/winml_unittests.cmake
+++ b/cmake/winml_unittests.cmake
@@ -4,7 +4,6 @@
 set(WINML_TEST_SRC_DIR ${REPO_ROOT}/winml/test)
 set(WINML_TEST_INC_DIR
   ${REPO_ROOT}/winml/test/common
-  ${REPO_ROOT}/winml/lib/Api.Image/inc
   ${REPO_ROOT}/winml/lib/Common/inc
   ${REPO_ROOT}/onnxruntime
   ${REPO_ROOT}/onnxruntime/core/providers/dml/DmlExecutionProvider/src/External/D3DX12
@@ -44,7 +43,7 @@ function(add_winml_test)
   if (_UT_DEPENDS)
     add_dependencies(${_UT_TARGET} ${_UT_DEPENDS})
   endif()
-  target_link_libraries(${_UT_TARGET} PRIVATE ${_UT_LIBS} gtest winml_lib_image ${onnxruntime_EXTERNAL_LIBRARIES} winml_lib_telemetry winml_lib_api onnxruntime)
+  target_link_libraries(${_UT_TARGET} PRIVATE ${_UT_LIBS} gtest ${onnxruntime_EXTERNAL_LIBRARIES} winml_lib_common onnxruntime)
 
   add_test(NAME ${_UT_TARGET}
     COMMAND ${_UT_TARGET}
diff --git a/winml/lib/Api.Image/D3DDeviceCache.cpp b/winml/lib/Api.Image/D3DDeviceCache.cpp
index a551bb96562c5..1b6314cd8a0bb 100644
--- a/winml/lib/Api.Image/D3DDeviceCache.cpp
+++ b/winml/lib/Api.Image/D3DDeviceCache.cpp
@@ -6,6 +6,7 @@
 #include <directxmath.h>
 #include <d3d11on12.h>
 #include "inc/DeviceHelpers.h"
+#include "CommonDeviceHelpers.h"
 
 namespace float32 {
 #include "shaders\SurfaceToTensor-SurfaceToTensorBGR8.h"
@@ -50,8 +51,8 @@ D3DDeviceCache::D3DDeviceCache(Windows::AI::MachineLearning::LearningModelDevice
   DXGI_GPU_PREFERENCE preference;
   WINML_THROW_IF_FAILED(DeviceHelpers::GetGPUPreference(deviceKind, &preference));
 
-  DeviceHelpers::AdapterEnumerationSupport support;
-  WINML_THROW_IF_FAILED(DeviceHelpers::GetAdapterEnumerationSupport(&support));
+  CommonDeviceHelpers::AdapterEnumerationSupport support;
+  WINML_THROW_IF_FAILED(CommonDeviceHelpers::GetAdapterEnumerationSupport(&support));
 
   const char errStr[] = "No hardware adapters available";
   if (support.has_dxgi) {
@@ -130,7 +131,7 @@ D3DDeviceCache::~D3DDeviceCache() {
 
 bool D3DDeviceCache::IsFloat16Supported() {
   if (device_ != nullptr) {
-    return DeviceHelpers::IsFloat16Supported(device_.get());
+    return CommonDeviceHelpers::IsFloat16Supported(device_.get());
   }
 
   return true;
@@ -671,4 +672,4 @@ void D3DDeviceCache::SyncD3D11DeviceToConverter(_In_ ID3D11Fence* pD3D11Fence) {
 bool D3DDeviceCache::SharedHandleInitialized() {
   return d3d11_fence_ != nullptr;
 }
-}  // namespace winrt::Windows::AI::MachineLearning::implementation
\ No newline at end of file
+}  // namespace winrt::Windows::AI::MachineLearning::implementation
diff --git a/winml/lib/Api.Image/DeviceHelpers.cpp b/winml/lib/Api.Image/DeviceHelpers.cpp
index 3174f61fad619..6e0b5a7ffd946 100644
--- a/winml/lib/Api.Image/DeviceHelpers.cpp
+++ b/winml/lib/Api.Image/DeviceHelpers.cpp
@@ -9,215 +9,10 @@
 #include <d3d11on12.h>
 #include <wil/winrt.h>
 #include "inc/DeviceHelpers.h"
+#include "CommonDeviceHelpers.h"
 #include "LearningModelDevice.h"
 
 namespace DeviceHelpers {
-constexpr uint32_t c_intelVendorId = 0x8086;
-constexpr uint32_t c_nvidiaVendorId = 0x10DE;
-constexpr uint32_t c_amdVendorId = 0x1002;
-static std::optional<AdapterEnumerationSupport> s_adapterEnumerationSupport;
-
-static bool CheckAdapterFP16Blocked(bool isMcdmAdapter, uint32_t vendorId, uint32_t majorVersion, uint32_t minorVersion) {
-  switch (vendorId) {
-    case c_intelVendorId: {
-      if (isMcdmAdapter) {
-        return false;
-      }
-
-      // Check Intel GPU driver version
-      return (majorVersion < 25) || (majorVersion == 25 && minorVersion < 6574) || (majorVersion == 26 && minorVersion < 6572);
-    }
-  }
-  return false;
-}
-
-static void ParseDriverVersion(LARGE_INTEGER& version, uint32_t& majorVersion, uint32_t& minorVersion) {
-  majorVersion = HIWORD(version.HighPart);
-  minorVersion = LOWORD(version.LowPart);
-}
-
-static HRESULT GetDXGIAdapterMetadata(ID3D12Device& device, uint32_t& vendorId, uint32_t& majorVersion, uint32_t& minorVersion) {
-  winrt::com_ptr<IDXGIFactory4> spFactory;
-  RETURN_IF_FAILED(CreateDXGIFactory1(IID_PPV_ARGS(spFactory.put())));
-
-  winrt::com_ptr<IDXGIAdapter> spAdapter;
-  RETURN_IF_FAILED(spFactory->EnumAdapterByLuid(device.GetAdapterLuid(), IID_PPV_ARGS(spAdapter.put())));
-
-  DXGI_ADAPTER_DESC adapterDesc = {};
-  RETURN_IF_FAILED(spAdapter->GetDesc(&adapterDesc));
-
-  LARGE_INTEGER driverVersion;
-  RETURN_IF_FAILED(spAdapter->CheckInterfaceSupport(__uuidof(IDXGIDevice), &driverVersion));
-
-  vendorId = adapterDesc.VendorId;
-  ParseDriverVersion(driverVersion, majorVersion, minorVersion);
-  return S_OK;
-}
-
-#ifdef ENABLE_DXCORE
-static HRESULT GetDXCoreAdapterMetadata(ID3D12Device& device, bool& isMcdmAdapter, uint32_t& vendorId, uint32_t& majorVersion, uint32_t& minorVersion) {
-  winrt::com_ptr<IDXCoreAdapterFactory> spFactory;
-  RETURN_IF_FAILED(DXCoreCreateAdapterFactory(IID_PPV_ARGS(spFactory.put())));
-
-  winrt::com_ptr<IDXCoreAdapter> spAdapter;
-  RETURN_IF_FAILED(spFactory->GetAdapterByLuid(device.GetAdapterLuid(), IID_PPV_ARGS(spAdapter.put())));
-
-  if (spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_CORE_COMPUTE) &&
-      (!(spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_GRAPHICS) ||
-         spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D11_GRAPHICS)))) {
-    isMcdmAdapter = true;
-  } else {
-    isMcdmAdapter = false;
-  }
-
-  DXCoreHardwareID hardwareId;
-  RETURN_IF_FAILED(spAdapter->GetProperty(DXCoreAdapterProperty::HardwareID, &hardwareId));
-  vendorId = hardwareId.vendorID;
-
-  uint64_t rawDriverVersion;
-  RETURN_IF_FAILED(spAdapter->GetProperty(DXCoreAdapterProperty::DriverVersion, &rawDriverVersion));
-
-  LARGE_INTEGER driverVersion;
-  driverVersion.QuadPart = static_cast<LONGLONG>(rawDriverVersion);
-  ParseDriverVersion(driverVersion, majorVersion, minorVersion);
-  return S_OK;
-}
-#endif
-
-static HRESULT GetD3D12Device(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device, ID3D12Device** outDevice) {
-  _LUID id;
-  id.LowPart = device.AdapterId().LowPart;
-  id.HighPart = device.AdapterId().HighPart;
-  AdapterEnumerationSupport support;
-  RETURN_IF_FAILED(GetAdapterEnumerationSupport(&support));
-
-  if (support.has_dxgi) {
-    winrt::com_ptr<IDXGIFactory6> spFactory;
-    RETURN_IF_FAILED(CreateDXGIFactory1(IID_PPV_ARGS(spFactory.put())));
-
-    winrt::com_ptr<IDXGIAdapter1> spAdapter;
-    RETURN_IF_FAILED(spFactory->EnumAdapterByLuid(id, IID_PPV_ARGS(spAdapter.put())));
-    RETURN_IF_FAILED(D3D12CreateDevice(spAdapter.get(), D3D_FEATURE_LEVEL_11_0, IID_PPV_ARGS(outDevice)));
-  }
-#ifdef ENABLE_DXCORE
-  if (support.has_dxgi == false) {
-    winrt::com_ptr<IDXCoreAdapterFactory> spFactory;
-    RETURN_IF_FAILED(DXCoreCreateAdapterFactory(IID_PPV_ARGS(spFactory.put())));
-
-    winrt::com_ptr<IDXCoreAdapter> spAdapter;
-    RETURN_IF_FAILED(spFactory->GetAdapterByLuid(id, IID_PPV_ARGS(spAdapter.put())));
-    RETURN_IF_FAILED(D3D12CreateDevice(spAdapter.get(), D3D_FEATURE_LEVEL_11_0, IID_PPV_ARGS(outDevice)));
-  }
-#endif
-  return S_OK;
-}
-
-static HRESULT IsFloat16Blocked(ID3D12Device& device, bool* isBlocked) {
-  uint32_t vendorId;
-  uint32_t majorVersion;
-  uint32_t minorVersion;
-  bool isMcdmAdapter;
-  *isBlocked = true;
-  AdapterEnumerationSupport support;
-  RETURN_IF_FAILED(GetAdapterEnumerationSupport(&support));
-#ifdef ENABLE_DXCORE
-  if (support.has_dxcore) {
-    RETURN_IF_FAILED(GetDXCoreAdapterMetadata(device, isMcdmAdapter, vendorId, majorVersion, minorVersion));
-    *isBlocked = CheckAdapterFP16Blocked(isMcdmAdapter, vendorId, majorVersion, minorVersion);
-    return S_OK;
-  }
-#endif
-  RETURN_IF_FAILED(GetDXGIAdapterMetadata(device, vendorId, majorVersion, minorVersion));
-  isMcdmAdapter = false;
-  *isBlocked = CheckAdapterFP16Blocked(isMcdmAdapter, vendorId, majorVersion, minorVersion);
-  return S_OK;
-}
-
-bool IsFloat16Supported(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device) {
-  auto modelImpl = device.as<winmlp::LearningModelDevice>();
-  if (modelImpl->IsCpuDevice()) {
-    return true;
-  }
-  winrt::com_ptr<ID3D12Device> d3d12Device;
-  if (FAILED(GetD3D12Device(device, d3d12Device.put()))) {
-    return false;
-  }
-  return IsFloat16Supported(d3d12Device.get());
-}
-
-bool IsFloat16Supported(ID3D12Device* device) {
-#ifndef USE_DML
-    WINML_THROW_HR_IF_TRUE_MSG(ERROR_NOT_SUPPORTED, true, "IsFloat16Supported is not implemented for WinML only build.");
-    return false;
-#else
-  bool isBlocked;
-  if (FAILED(IsFloat16Blocked(*device, &isBlocked)) || isBlocked) {
-    return false;
-  }
-  winrt::com_ptr<IDMLDevice> dmlDevice;
-  winrt::check_hresult(DMLCreateDevice(
-      device,
-      DML_CREATE_DEVICE_FLAG_NONE,
-      IID_PPV_ARGS(dmlDevice.put())));
-
-  DML_FEATURE_QUERY_TENSOR_DATA_TYPE_SUPPORT float16Query = {DML_TENSOR_DATA_TYPE_FLOAT16};
-  DML_FEATURE_DATA_TENSOR_DATA_TYPE_SUPPORT float16Data = {};
-
-  winrt::check_hresult(dmlDevice->CheckFeatureSupport(
-      DML_FEATURE_TENSOR_DATA_TYPE_SUPPORT,
-      sizeof(float16Query),
-      &float16Query,
-      sizeof(float16Data),
-      &float16Data));
-  return float16Data.IsSupported;
-#endif USE_DML
-}
-
-// uses Structured Exception Handling (SEH) to detect for delay load failures of target API.
-// You cannot mix and match SEH with C++ exception and object unwinding
-// In this case we will catch it, and report up to the caller via HRESULT so our callers can use
-// C++ exceptions
-template <typename TFunc, typename... TArgs>
-static HRESULT RunDelayLoadedApi(TFunc& tfunc, TArgs&&... args) {
-  __try {
-    return tfunc(std::forward<TArgs>(args)...);
-  } __except (GetExceptionCode() == VcppException(ERROR_SEVERITY_ERROR, ERROR_MOD_NOT_FOUND) ? EXCEPTION_EXECUTE_HANDLER : EXCEPTION_CONTINUE_SEARCH) {
-    // this could be ok, just let people know that it failed to load
-    return HRESULT_FROM_WIN32(ERROR_MOD_NOT_FOUND);
-  }
-}
-
-HRESULT GetAdapterEnumerationSupport(AdapterEnumerationSupport* support) {
-  if (!s_adapterEnumerationSupport.has_value()) {
-    // check for support, starting with DXGI
-    winrt::com_ptr<IDXGIFactory4> dxgiFactory;
-#ifdef ENABLE_DXCORE
-    winrt::com_ptr<IDXCoreAdapterFactory> dxcoreFactory;
-    // necessary because DXCoreCreateAdapterFactory is overloaded
-    HRESULT(WINAPI * pDxCoreTestFunc)
-    (REFIID, void**) = DXCoreCreateAdapterFactory;
-#endif
-    AdapterEnumerationSupport adapterEnumerationSupport = {};
-
-    if (SUCCEEDED(RunDelayLoadedApi(CreateDXGIFactory1, IID_PPV_ARGS(dxgiFactory.put())))) {
-      adapterEnumerationSupport.has_dxgi = true;
-    }
-#ifdef ENABLE_DXCORE
-    if (SUCCEEDED(RunDelayLoadedApi(pDxCoreTestFunc, IID_PPV_ARGS(dxcoreFactory.put())))) {
-      adapterEnumerationSupport.has_dxcore = true;
-    }
-#endif
-
-    s_adapterEnumerationSupport = adapterEnumerationSupport;
-
-    if (!(adapterEnumerationSupport.has_dxgi || adapterEnumerationSupport.has_dxcore)) {
-      return TYPE_E_CANTLOADLIBRARY;
-    }
-  }
-  *support = s_adapterEnumerationSupport.value();
-  return S_OK;
-}
-
 HRESULT GetDXGIHardwareAdapterWithPreference(DXGI_GPU_PREFERENCE preference, IDXGIAdapter1** ppAdapter) {
   winrt::com_ptr<IDXGIFactory6> spFactory;
   RETURN_IF_FAILED(CreateDXGIFactory1(IID_PPV_ARGS(spFactory.put())));
@@ -304,7 +99,7 @@ HRESULT GetDXCoreHardwareAdapterWithPreference(DXGI_GPU_PREFERENCE preference, I
 #endif
 
 HRESULT CreateD3D11On12Device(ID3D12Device* device12, ID3D11Device** device11) {
-  return DeviceHelpers::RunDelayLoadedApi(
+  return CommonDeviceHelpers::RunDelayLoadedApi(
       D3D11On12CreateDevice,
       device12,                          // pointer to d3d12 device
       D3D11_CREATE_DEVICE_BGRA_SUPPORT,  // required in order to interop with Direct2D
diff --git a/winml/lib/Api.Image/inc/DeviceHelpers.h b/winml/lib/Api.Image/inc/DeviceHelpers.h
index 2a86c7d40cd96..d5c409c02b680 100644
--- a/winml/lib/Api.Image/inc/DeviceHelpers.h
+++ b/winml/lib/Api.Image/inc/DeviceHelpers.h
@@ -14,28 +14,11 @@
 #include <dxcore.h>
 #endif
 
-//
-// Exception information
-//
-#ifndef FACILITY_VISUALCPP
-#define FACILITY_VISUALCPP ((LONG)0x6d)
-#endif
-
-#define VcppException(sev, err) ((sev) | (FACILITY_VISUALCPP << 16) | err)
-
 namespace DeviceHelpers {
-struct AdapterEnumerationSupport {
-  bool has_dxgi;
-  bool has_dxcore;
-};
-
-HRESULT GetAdapterEnumerationSupport(AdapterEnumerationSupport* support);
-bool IsFloat16Supported(ID3D12Device* device);
-bool IsFloat16Supported(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device);
 HRESULT CreateD3D11On12Device(ID3D12Device* device12, ID3D11Device** device11);
 #ifdef ENABLE_DXCORE
 HRESULT GetDXCoreHardwareAdapterWithPreference(DXGI_GPU_PREFERENCE preference, _COM_Outptr_ IDXCoreAdapter** ppAdapter);
 #endif
 HRESULT GetDXGIHardwareAdapterWithPreference(DXGI_GPU_PREFERENCE preference, _COM_Outptr_ IDXGIAdapter1** adapter);
 HRESULT GetGPUPreference(winrt::Windows::AI::MachineLearning::LearningModelDeviceKind deviceKind, DXGI_GPU_PREFERENCE* preference) noexcept;
-}  // namespace DeviceHelpers
\ No newline at end of file
+}  // namespace DeviceHelpers
diff --git a/winml/lib/Common/CommonDeviceHelpers.cpp b/winml/lib/Common/CommonDeviceHelpers.cpp
new file mode 100644
index 0000000000000..292d665ffdf1b
--- /dev/null
+++ b/winml/lib/Common/CommonDeviceHelpers.cpp
@@ -0,0 +1,215 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT License.
+
+// #include "dx.h"
+// #include <D3dx12.h>
+
+#if USE_DML
+#include <DirectML.h>
+#endif USE_DML
+#include "inc/CommonDeviceHelpers.h"
+#include <d3d11on12.h>
+#include <wil/winrt.h>
+#include "LearningModelDevice.h"
+
+namespace {
+constexpr uint32_t c_intelVendorId = 0x8086;
+constexpr uint32_t c_nvidiaVendorId = 0x10DE;
+constexpr uint32_t c_amdVendorId = 0x1002;
+
+bool CheckAdapterFP16Blocked(bool isMcdmAdapter, uint32_t vendorId, uint32_t majorVersion, uint32_t minorVersion) {
+  switch (vendorId) {
+    case c_intelVendorId: {
+      if (isMcdmAdapter) {
+        return false;
+      }
+
+      // Check Intel GPU driver version
+      return (majorVersion < 25) || (majorVersion == 25 && minorVersion < 6574) || (majorVersion == 26 && minorVersion < 6572);
+    }
+  }
+  return false;
+}
+
+void ParseDriverVersion(LARGE_INTEGER& version, uint32_t& majorVersion, uint32_t& minorVersion) {
+  majorVersion = HIWORD(version.HighPart);
+  minorVersion = LOWORD(version.LowPart);
+}
+
+HRESULT GetDXGIAdapterMetadata(ID3D12Device& device, uint32_t& vendorId, uint32_t& majorVersion, uint32_t& minorVersion) {
+  winrt::com_ptr<IDXGIFactory4> spFactory;
+  RETURN_IF_FAILED(CreateDXGIFactory1(IID_PPV_ARGS(spFactory.put())));
+
+  winrt::com_ptr<IDXGIAdapter> spAdapter;
+  RETURN_IF_FAILED(spFactory->EnumAdapterByLuid(device.GetAdapterLuid(), IID_PPV_ARGS(spAdapter.put())));
+
+  DXGI_ADAPTER_DESC adapterDesc = {};
+  RETURN_IF_FAILED(spAdapter->GetDesc(&adapterDesc));
+
+  LARGE_INTEGER driverVersion;
+  RETURN_IF_FAILED(spAdapter->CheckInterfaceSupport(__uuidof(IDXGIDevice), &driverVersion));
+
+  vendorId = adapterDesc.VendorId;
+  ParseDriverVersion(driverVersion, majorVersion, minorVersion);
+  return S_OK;
+}
+
+#ifdef ENABLE_DXCORE
+HRESULT GetDXCoreAdapterMetadata(ID3D12Device& device, bool& isMcdmAdapter, uint32_t& vendorId, uint32_t& majorVersion, uint32_t& minorVersion) {
+  winrt::com_ptr<IDXCoreAdapterFactory> spFactory;
+  RETURN_IF_FAILED(DXCoreCreateAdapterFactory(IID_PPV_ARGS(spFactory.put())));
+
+  winrt::com_ptr<IDXCoreAdapter> spAdapter;
+  RETURN_IF_FAILED(spFactory->GetAdapterByLuid(device.GetAdapterLuid(), IID_PPV_ARGS(spAdapter.put())));
+
+  if (spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_CORE_COMPUTE) &&
+      (!(spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_GRAPHICS) ||
+         spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D11_GRAPHICS)))) {
+    isMcdmAdapter = true;
+  } else {
+    isMcdmAdapter = false;
+  }
+
+  DXCoreHardwareID hardwareId;
+  RETURN_IF_FAILED(spAdapter->GetProperty(DXCoreAdapterProperty::HardwareID, &hardwareId));
+  vendorId = hardwareId.vendorID;
+
+  uint64_t rawDriverVersion;
+  RETURN_IF_FAILED(spAdapter->GetProperty(DXCoreAdapterProperty::DriverVersion, &rawDriverVersion));
+
+  LARGE_INTEGER driverVersion;
+  driverVersion.QuadPart = static_cast<LONGLONG>(rawDriverVersion);
+  ParseDriverVersion(driverVersion, majorVersion, minorVersion);
+  return S_OK;
+}
+#endif
+
+HRESULT GetD3D12Device(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device, ID3D12Device** outDevice) {
+  _LUID id;
+  id.LowPart = device.AdapterId().LowPart;
+  id.HighPart = device.AdapterId().HighPart;
+  CommonDeviceHelpers::AdapterEnumerationSupport support;
+  RETURN_IF_FAILED(GetAdapterEnumerationSupport(&support));
+
+  if (support.has_dxgi) {
+    winrt::com_ptr<IDXGIFactory6> spFactory;
+    RETURN_IF_FAILED(CreateDXGIFactory1(IID_PPV_ARGS(spFactory.put())));
+
+    winrt::com_ptr<IDXGIAdapter1> spAdapter;
+    RETURN_IF_FAILED(spFactory->EnumAdapterByLuid(id, IID_PPV_ARGS(spAdapter.put())));
+    RETURN_IF_FAILED(D3D12CreateDevice(spAdapter.get(), D3D_FEATURE_LEVEL_11_0, IID_PPV_ARGS(outDevice)));
+  }
+#ifdef ENABLE_DXCORE
+  if (support.has_dxgi == false) {
+    winrt::com_ptr<IDXCoreAdapterFactory> spFactory;
+    RETURN_IF_FAILED(DXCoreCreateAdapterFactory(IID_PPV_ARGS(spFactory.put())));
+
+    winrt::com_ptr<IDXCoreAdapter> spAdapter;
+    RETURN_IF_FAILED(spFactory->GetAdapterByLuid(id, IID_PPV_ARGS(spAdapter.put())));
+    RETURN_IF_FAILED(D3D12CreateDevice(spAdapter.get(), D3D_FEATURE_LEVEL_11_0, IID_PPV_ARGS(outDevice)));
+  }
+#endif
+  return S_OK;
+}
+
+HRESULT IsFloat16Blocked(ID3D12Device& device, bool* isBlocked) {
+  uint32_t vendorId;
+  uint32_t majorVersion;
+  uint32_t minorVersion;
+  bool isMcdmAdapter;
+  *isBlocked = true;
+  CommonDeviceHelpers::AdapterEnumerationSupport support;
+  RETURN_IF_FAILED(CommonDeviceHelpers::GetAdapterEnumerationSupport(&support));
+#ifdef ENABLE_DXCORE
+  if (support.has_dxcore) {
+    RETURN_IF_FAILED(GetDXCoreAdapterMetadata(device, isMcdmAdapter, vendorId, majorVersion, minorVersion));
+    *isBlocked = CheckAdapterFP16Blocked(isMcdmAdapter, vendorId, majorVersion, minorVersion);
+    return S_OK;
+  }
+#endif
+  RETURN_IF_FAILED(GetDXGIAdapterMetadata(device, vendorId, majorVersion, minorVersion));
+  isMcdmAdapter = false;
+  *isBlocked = CheckAdapterFP16Blocked(isMcdmAdapter, vendorId, majorVersion, minorVersion);
+  return S_OK;
+}
+}
+
+namespace CommonDeviceHelpers {
+constexpr uint32_t c_intelVendorId = 0x8086;
+constexpr uint32_t c_nvidiaVendorId = 0x10DE;
+constexpr uint32_t c_amdVendorId = 0x1002;
+
+bool IsFloat16Supported(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device) {
+  auto adapterId = device.AdapterId();
+  if (!adapterId.HighPart && !adapterId.LowPart) {
+    // CPU device
+    return true;
+  }
+  winrt::com_ptr<ID3D12Device> d3d12Device;
+  if (FAILED(GetD3D12Device(device, d3d12Device.put()))) {
+    return false;
+  }
+  return IsFloat16Supported(d3d12Device.get());
+}
+
+bool IsFloat16Supported(ID3D12Device* device) {
+#ifndef USE_DML
+    WINML_THROW_HR_IF_TRUE_MSG(ERROR_NOT_SUPPORTED, true, "IsFloat16Supported is not implemented for WinML only build.");
+    return false;
+#else
+  bool isBlocked;
+  if (FAILED(IsFloat16Blocked(*device, &isBlocked)) || isBlocked) {
+    return false;
+  }
+  winrt::com_ptr<IDMLDevice> dmlDevice;
+  winrt::check_hresult(DMLCreateDevice(
+      device,
+      DML_CREATE_DEVICE_FLAG_NONE,
+      IID_PPV_ARGS(dmlDevice.put())));
+
+  DML_FEATURE_QUERY_TENSOR_DATA_TYPE_SUPPORT float16Query = {DML_TENSOR_DATA_TYPE_FLOAT16};
+  DML_FEATURE_DATA_TENSOR_DATA_TYPE_SUPPORT float16Data = {};
+
+  winrt::check_hresult(dmlDevice->CheckFeatureSupport(
+      DML_FEATURE_TENSOR_DATA_TYPE_SUPPORT,
+      sizeof(float16Query),
+      &float16Query,
+      sizeof(float16Data),
+      &float16Data));
+  return float16Data.IsSupported;
+#endif
+}
+
+HRESULT GetAdapterEnumerationSupport(AdapterEnumerationSupport* support) {
+  static std::optional<AdapterEnumerationSupport> s_adapterEnumerationSupport;
+  if (!s_adapterEnumerationSupport.has_value()) {
+    // check for support, starting with DXGI
+    winrt::com_ptr<IDXGIFactory4> dxgiFactory;
+#ifdef ENABLE_DXCORE
+    winrt::com_ptr<IDXCoreAdapterFactory> dxcoreFactory;
+    // necessary because DXCoreCreateAdapterFactory is overloaded
+    HRESULT(WINAPI * pDxCoreTestFunc)
+    (REFIID, void**) = DXCoreCreateAdapterFactory;
+#endif
+    AdapterEnumerationSupport adapterEnumerationSupport = {};
+
+    if (SUCCEEDED(RunDelayLoadedApi(CreateDXGIFactory1, IID_PPV_ARGS(dxgiFactory.put())))) {
+      adapterEnumerationSupport.has_dxgi = true;
+    }
+#ifdef ENABLE_DXCORE
+    if (SUCCEEDED(RunDelayLoadedApi(pDxCoreTestFunc, IID_PPV_ARGS(dxcoreFactory.put())))) {
+      adapterEnumerationSupport.has_dxcore = true;
+    }
+#endif
+
+    s_adapterEnumerationSupport = adapterEnumerationSupport;
+
+    if (!(adapterEnumerationSupport.has_dxgi || adapterEnumerationSupport.has_dxcore)) {
+      return TYPE_E_CANTLOADLIBRARY;
+    }
+  }
+  *support = s_adapterEnumerationSupport.value();
+  return S_OK;
+}
+
+}  // namespace CommonDeviceHelpers
diff --git a/winml/lib/Common/inc/CommonDeviceHelpers.h b/winml/lib/Common/inc/CommonDeviceHelpers.h
new file mode 100644
index 0000000000000..38f1aa1e8a13c
--- /dev/null
+++ b/winml/lib/Common/inc/CommonDeviceHelpers.h
@@ -0,0 +1,50 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT License.
+
+#pragma once
+
+#include "winrt_headers.h"
+#include <dxgi1_6.h>
+#include <initguid.h>
+#include <d3d11.h>
+
+#if __has_include("dxcore.h")
+#define ENABLE_DXCORE 1
+#endif
+#ifdef ENABLE_DXCORE
+#include <dxcore.h>
+#endif
+
+//
+// Exception information
+//
+#ifndef FACILITY_VISUALCPP
+#define FACILITY_VISUALCPP ((LONG)0x6d)
+#endif
+
+#define VcppException(sev, err) ((sev) | (FACILITY_VISUALCPP << 16) | err)
+
+namespace CommonDeviceHelpers {
+struct AdapterEnumerationSupport {
+  bool has_dxgi;
+  bool has_dxcore;
+};
+
+// uses Structured Exception Handling (SEH) to detect for delay load failures of target API.
+// You cannot mix and match SEH with C++ exception and object unwinding
+// In this case we will catch it, and report up to the caller via HRESULT so our callers can use
+// C++ exceptions
+template <typename TFunc, typename... TArgs>
+HRESULT RunDelayLoadedApi(TFunc& tfunc, TArgs&&... args) {
+  __try {
+    return tfunc(std::forward<TArgs>(args)...);
+  } __except (GetExceptionCode() == VcppException(ERROR_SEVERITY_ERROR, ERROR_MOD_NOT_FOUND) ? EXCEPTION_EXECUTE_HANDLER : EXCEPTION_CONTINUE_SEARCH) {
+    // this could be ok, just let people know that it failed to load
+    return HRESULT_FROM_WIN32(ERROR_MOD_NOT_FOUND);
+  }
+}
+
+HRESULT GetAdapterEnumerationSupport(AdapterEnumerationSupport* support);
+bool IsFloat16Supported(ID3D12Device* device);
+bool IsFloat16Supported(const winrt::Windows::AI::MachineLearning::LearningModelDevice& device);
+}  // namespace CommonDeviceHelpers
diff --git a/winml/test/api/LearningModelBindingAPITest.cpp b/winml/test/api/LearningModelBindingAPITest.cpp
index 99d9dc6e9b251..883eeb3cfc326 100644
--- a/winml/test/api/LearningModelBindingAPITest.cpp
+++ b/winml/test/api/LearningModelBindingAPITest.cpp
@@ -7,7 +7,6 @@
 #include <winrt/Windows.Graphics.Imaging.h>
 #include <winrt/Windows.Media.h>
 #include "winrt/Windows.Storage.h"
-#include "DeviceHelpers.h"
 #include <sstream>
 using namespace winrt;
 using namespace winrt::Windows::AI::MachineLearning;
@@ -655,4 +654,4 @@ const LearningModelBindingAPITestApi& getapi() {
     VerifyOutputAfterImageBindCalledTwice
   };
   return api;
-}
\ No newline at end of file
+}
diff --git a/winml/test/api/LearningModelSessionAPITest.cpp b/winml/test/api/LearningModelSessionAPITest.cpp
index 908d13532e44d..45af7cbd831b1 100644
--- a/winml/test/api/LearningModelSessionAPITest.cpp
+++ b/winml/test/api/LearningModelSessionAPITest.cpp
@@ -1,10 +1,10 @@
 #include "testPch.h"
+
 #include "APITest.h"
+#include "CommonDeviceHelpers.h"
 #include "LearningModelSessionAPITest.h"
-#include "winrt/Windows.Storage.h"
-
-#include "DeviceHelpers.h"
 #include "protobufHelpers.h"
+#include "winrt/Windows.Storage.h"
 
 #include <D3d11_4.h>
 #include <dxgi1_6.h>
@@ -231,7 +231,7 @@ static LearningModelSession CreateSession(LearningModel model)
     WINML_EXPECT_NO_THROW(device = LearningModelDevice(LearningModelDeviceKind::DirectX));
 
     LearningModelSession session(nullptr);
-    if (DeviceHelpers::IsFloat16Supported(device))
+    if (CommonDeviceHelpers::IsFloat16Supported(device))
     {
         WINML_EXPECT_NO_THROW(session = LearningModelSession(model, device));
     }
diff --git a/winml/test/common/std.h b/winml/test/common/std.h
index d6eab75f645f8..99a3af8e5943f 100644
--- a/winml/test/common/std.h
+++ b/winml/test/common/std.h
@@ -24,9 +24,10 @@
 // IUnknown must be declared before winrt/base.h is included to light up support for native COM
 // interfaces with C++/WinRT types (e.g. winrt::com_ptr<ITensorNative>).
 #include <Unknwn.h>
+#include <wil/cppwinrt.h>
 #include "winrt/base.h"
 #include "winrt/Windows.Foundation.Collections.h"
 #include "comp_generated/winrt/windows.ai.machinelearning.h"
 
 // WinML
-#include "Windows.AI.MachineLearning.Native.h"
\ No newline at end of file
+#include "Windows.AI.MachineLearning.Native.h"
diff --git a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
index 1420b0c05e425..e38f5c29b3079 100644
--- a/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
+++ b/winml/test/scenario/cppwinrt/scenariotestscppwinrt.cpp
@@ -15,20 +15,21 @@
 #ifdef GetCurrentTime
 #undef GetCurrentTime
 #endif
+#include "CommonDeviceHelpers.h"
 #include "CustomOperatorProvider.h"
-#include "DeviceHelpers.h"
 #include "filehelpers.h"
 #include "robuffer.h"
+#include "scenariotestscppwinrt.h"
 #include "Windows.AI.MachineLearning.Native.h"
 #include "Windows.Graphics.DirectX.Direct3D11.interop.h"
 #include "windows.ui.xaml.media.dxinterop.h"
 #include "winrt/Windows.UI.Xaml.Controls.h"
 #include "winrt/Windows.UI.Xaml.Media.Imaging.h"
+
 #include <d2d1.h>
 #include <d3d11.h>
 #include <initguid.h>
 #include <MemoryBuffer.h>
-#include "scenariotestscppwinrt.h"
 #include <iostream>
 #if __has_include("dxcore.h")
 #define ENABLE_DXCORE 1
@@ -408,8 +409,8 @@ static void Scenario8SetDeviceSampleCustomCommandQueue() {
   LearningModel model = LearningModel::LoadFromFilePath(filePath);
 
   com_ptr<ID3D12Device> pD3D12Device = nullptr;
-  DeviceHelpers::AdapterEnumerationSupport support;
-  if (FAILED(DeviceHelpers::GetAdapterEnumerationSupport(&support))) {
+  CommonDeviceHelpers::AdapterEnumerationSupport support;
+  if (FAILED(CommonDeviceHelpers::GetAdapterEnumerationSupport(&support))) {
     WINML_LOG_ERROR("Unable to load DXGI or DXCore");
     return;
   }
@@ -696,7 +697,7 @@ static void Scenario17DevDiagnostics() {
   WINML_EXPECT_NO_THROW(session.Evaluate(binding, L""));
 }
 
-/** 
+/**
  * Custom Operator Tests are labeled as GPU tests because the DML code is interlaced with the custom op code
  * even though CPU custom ops shouldn't be dependent on GPU functionality.
  * These should be reclassed to ScenarioCppWinrt once the DML code is decoupled from the custom op code.
@@ -1242,8 +1243,8 @@ static void ReuseVideoFrame() {
 
   std::vector<LearningModelDeviceKind> deviceKinds = {LearningModelDeviceKind::Cpu, LearningModelDeviceKind::DirectX};
   std::vector<std::string> videoFrameSources;
-  DeviceHelpers::AdapterEnumerationSupport support;
-  DeviceHelpers::GetAdapterEnumerationSupport(&support);
+  CommonDeviceHelpers::AdapterEnumerationSupport support;
+  CommonDeviceHelpers::GetAdapterEnumerationSupport(&support);
   if (support.has_dxgi) {
     videoFrameSources = {"SoftwareBitmap", "Direct3DSurface"};
   } else {
@@ -1449,4 +1450,4 @@ const ScenarioTestApi& getapi() {
         D2DInterop,
       };
   return api;
-}
\ No newline at end of file
+}

From 93f6e3e1161ab41782ea42f192e5d55c9d898712 Mon Sep 17 00:00:00 2001
From: Tiago Koji Castro Shibata <ticastro@microsoft.com>
Date: Fri, 17 Jan 2020 16:15:37 -0800
Subject: [PATCH 74/75] Support custom DML in onnxruntime_providers.cmake
 (#2867)

---
 cmake/onnxruntime_providers.cmake | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/cmake/onnxruntime_providers.cmake b/cmake/onnxruntime_providers.cmake
index 1b7047318ee85..3bf0b6859f74e 100644
--- a/cmake/onnxruntime_providers.cmake
+++ b/cmake/onnxruntime_providers.cmake
@@ -431,19 +431,23 @@ if (onnxruntime_USE_DML)
   add_dependencies(onnxruntime_providers_dml ${onnxruntime_EXTERNAL_DEPENDENCIES})
   target_include_directories(onnxruntime_providers_dml PRIVATE ${ONNXRUNTIME_ROOT} ${ONNXRUNTIME_ROOT}/../cmake/external/wil/include)
 
-  if(NOT onnxruntime_target_platform STREQUAL "x86" AND NOT onnxruntime_target_platform STREQUAL "x64")
-    message(FATAL_ERROR "Target platform ${onnxruntime_target_platform} is not supported by DML")
+  if (NOT onnxruntime_USE_CUSTOM_DIRECTML)
+    if(NOT onnxruntime_target_platform STREQUAL "x86" AND NOT onnxruntime_target_platform STREQUAL "x64")
+      message(FATAL_ERROR "Target platform ${onnxruntime_target_platform} is not supported by DML")
+    endif()
+    foreach(file "DirectML.dll" "DirectML.pdb" "DirectML.Debug.dll" "DirectML.Debug.pdb")
+      add_custom_command(TARGET onnxruntime_providers_dml
+        POST_BUILD
+        COMMAND ${CMAKE_COMMAND} -E copy_if_different
+          "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/${file}" $<TARGET_FILE_DIR:onnxruntime_providers_dml>)
+    endforeach()
   endif()
-  foreach(file "DirectML.dll" "DirectML.pdb" "DirectML.Debug.dll" "DirectML.Debug.pdb")
-    add_custom_command(TARGET onnxruntime_providers_dml
-      POST_BUILD
-      COMMAND ${CMAKE_COMMAND} -E copy_if_different
-        "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/${file}" $<TARGET_FILE_DIR:onnxruntime_providers_dml>)
-  endforeach()
 
   function(target_add_dml target)
-    target_link_libraries(${target} PRIVATE "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/DirectML.lib")
-    target_include_directories(${target} PRIVATE "${DML_PACKAGE_DIR}/include")
+    if (NOT onnxruntime_USE_CUSTOM_DIRECTML)
+      target_link_libraries(${target} PRIVATE "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/DirectML.lib")
+      target_include_directories(${target} PRIVATE "${DML_PACKAGE_DIR}/include")
+    endif()
   endfunction()
 
   target_add_dml(onnxruntime_providers_dml)

From 8fefe0134c7cbf94adca3d9eeafc36bba818311c Mon Sep 17 00:00:00 2001
From: Xiang Zhang <xianz@microsoft.com>
Date: Tue, 21 Jan 2020 15:45:59 -0800
Subject: [PATCH 75/75] remove old winmladapter cpp

---
 winml/adapter/WinMLAdapter.cpp | 759 ---------------------------------
 1 file changed, 759 deletions(-)
 delete mode 100644 winml/adapter/WinMLAdapter.cpp

diff --git a/winml/adapter/WinMLAdapter.cpp b/winml/adapter/WinMLAdapter.cpp
deleted file mode 100644
index eef56c60799a8..0000000000000
--- a/winml/adapter/WinMLAdapter.cpp
+++ /dev/null
@@ -1,759 +0,0 @@
-// Copyright (c) Microsoft Corporation.
-// Licensed under the MIT License.
-
-#include "pch.h"
-#include "WinMLAdapter.h"
-#include "WinMLAdapterErrors.h"
-#include "CustomRegistryHelper.h"
-#include "PheonixSingleton.h"
-#include "LotusEnvironment.h"
-#include "AbiCustomRegistryImpl.h"
-
-#ifdef USE_DML
-#include "core/providers/dml/DmlExecutionProvider/inc/DmlExecutionProvider.h"
-#include "core/providers/dml/GraphTransformers/GraphTransformerHelpers.h"
-#include "core/providers/dml/OperatorAuthorHelper/SchemaInferenceOverrider.h"
-#include "DmlOrtSessionBuilder.h"
-#endif USE_DML
-
-#include "LearningModelDevice.h"
-#include "TensorFeatureDescriptor.h"
-#include "ImageFeatureDescriptor.h"
-#include "api.image/inc/D3DDeviceCache.h"
-#include "Common/inc/WinMLTelemetryHelper.h"
-
-#include "CpuOrtSessionBuilder.h"
-
-#include <io.h>
-#include <fcntl.h>
-
-#include "ZeroCopyInputStreamWrapper.h"
-#include "google/protobuf/io/zero_copy_stream_impl.h"
-
-#include "FeatureDescriptorFactory.h"
-#include "core\framework\utils.h"
-#include "core\framework\session_state.h"
-#include "core/providers/winml/winml_provider_factory.h"
-
-using namespace winrt::Windows::AI::MachineLearning;
-
-namespace Windows::AI::MachineLearning::Adapter {
-
-// Define winml trace logging provider with WinML GUID
-TRACELOGGING_DEFINE_PROVIDER(
-    winml_trace_logging_provider,
-    WINML_PROVIDER_DESC,
-    WINML_PROVIDER_GUID);
-
-// ORT intentionally requires callers derive from their session class to access
-// the protected methods used below.
-class InferenceSessionProtectedLoadAccessor : public onnxruntime::InferenceSession {
- public:
-  onnxruntime::common::Status
-  Load(std::unique_ptr<ONNX_NAMESPACE::ModelProto> p_model_proto) {
-    return onnxruntime::InferenceSession::Load(std::move(p_model_proto));
-  }
-  const onnxruntime::SessionState& GetSessionState() {
-    return *session_state_;
-  }
-};
-
-class ModelProto : public Microsoft::WRL::RuntimeClass<
-                       Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>,
-                       IModelProto> {
- public:
-  ModelProto::ModelProto(onnx::ModelProto* model_proto) : model_proto_(model_proto) {
-  }
-
-  onnx::ModelProto* STDMETHODCALLTYPE get() noexcept override {
-    return model_proto_.get();
-  }
-
-  onnx::ModelProto* STDMETHODCALLTYPE detach() noexcept override {
-    return model_proto_.release();
-  }
-
- private:
-  std::unique_ptr<onnx::ModelProto> model_proto_;
-};  // class ModelProto
-
-class ModelInfo : public Microsoft::WRL::RuntimeClass<
-                      Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>,
-                      IModelInfo> {
- private:
-  std::string author_;
-  std::string name_;
-  std::string domain_;
-  std::string description_;
-  int64_t version_;
-  std::unordered_map<std::string, std::string> model_metadata_;
-  wfc::IVector<winml::ILearningModelFeatureDescriptor> input_features_;
-  wfc::IVector<winml::ILearningModelFeatureDescriptor> output_features_;
-
- public:
-  ModelInfo(const onnx::ModelProto* model_proto) {
-    Initialize(model_proto);
-  }
-
-  const char* STDMETHODCALLTYPE author() noexcept override {
-    return author_.c_str();
-  }
-
-  const char* STDMETHODCALLTYPE name() noexcept override {
-    return name_.c_str();
-  }
-
-  const char* STDMETHODCALLTYPE domain() noexcept override {
-    return domain_.c_str();
-  }
-
-  const char* STDMETHODCALLTYPE description() noexcept override {
-    return description_.c_str();
-  }
-
-  int64_t STDMETHODCALLTYPE version() noexcept override {
-    return version_;
-  }
-
-  HRESULT STDMETHODCALLTYPE GetModelMetadata(
-      ABI::Windows::Foundation::Collections::IMapView<HSTRING, HSTRING>** metadata) override try {
-    *metadata = nullptr;
-    std::unordered_map<winrt::hstring, winrt::hstring> map_copy;
-    for (auto& pair : model_metadata_) {
-      auto key = WinML::Strings::HStringFromUTF8(pair.first);
-      auto map_value = WinML::Strings::HStringFromUTF8(pair.second);
-      map_copy.emplace(std::move(key), std::move(map_value));
-    }
-    auto out = winrt::single_threaded_map<winrt::hstring, winrt::hstring>(
-        std::move(map_copy));
-
-    winrt::copy_to_abi(out.GetView(), *(void**)metadata);
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetInputFeatures(
-      ABI::Windows::Foundation::Collections::IVectorView<winml::ILearningModelFeatureDescriptor>** features) override try {
-    *features = nullptr;
-    winrt::copy_to_abi(input_features_.GetView(), *(void**)features);
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetOutputFeatures(
-      ABI::Windows::Foundation::Collections::IVectorView<winml::ILearningModelFeatureDescriptor>** features) override try {
-    *features = nullptr;
-    winrt::copy_to_abi(output_features_.GetView(), *(void**)features);
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  static std::vector<const char*>
-  GetAllNodeOutputs(const onnx::ModelProto& model_proto) {
-    std::vector<const char*> nodes_outputs;
-    auto& graph = model_proto.graph();
-    auto& nodes = graph.node();
-    for (auto& node : nodes) {
-      for (auto& node_output : node.output()) {
-        nodes_outputs.push_back(node_output.c_str());
-      }
-    }
-    return nodes_outputs;
-  }
-
-  static std::vector<const char*>
-  GetInitializers(const onnx::ModelProto& model_proto) {
-    std::vector<const char*> initializers;
-    auto& graph = model_proto.graph();
-    auto& graph_initializers = graph.initializer();
-    for (auto& initializer : graph_initializers) {
-      initializers.push_back(initializer.name().c_str());
-    }
-    return initializers;
-  }
-
-  static std::vector<const onnx::ValueInfoProto*>
-  GetInputsWithoutInitializers(const onnx::ModelProto& model_proto) {
-    auto initializers = GetInitializers(model_proto);
-
-    std::vector<const onnx::ValueInfoProto*> inputs_without_initializers;
-    auto& graph = model_proto.graph();
-    auto& inputs = graph.input();
-    for (auto& input : inputs) {
-      if (input.has_name() && input.has_type()) {
-        auto found_it = std::find_if(
-            std::begin(initializers),
-            std::end(initializers),
-            [&](auto& initializer) {
-              return std::strcmp(initializer, input.name().c_str()) == 0;
-            });
-
-        auto is_initializer = found_it != std::end(initializers);
-        if (!is_initializer) {
-          inputs_without_initializers.push_back(&input);
-        }
-      }
-    }
-    return inputs_without_initializers;
-  }
-
-  static std::vector<const onnx::ValueInfoProto*> GetOutputs(const onnx::ModelProto& model_proto) {
-    std::vector<const onnx::ValueInfoProto*> outputs_with_name;
-    auto& graph = model_proto.graph();
-    auto& outputs = graph.output();
-    for (auto& output : outputs) {
-      if (output.has_name() && output.has_type()) {
-        outputs_with_name.push_back(&output);
-      }
-    }
-    return outputs_with_name;
-  }
-
- private:
-  void Initialize(const onnx::ModelProto* model_proto) {
-    // metadata
-    for (auto& prop : model_proto->metadata_props()) {
-      model_metadata_[prop.key()] = prop.value();
-    }
-
-    WinML::FeatureDescriptorFactory builder(model_metadata_);
-
-    // Create inputs
-    auto inputs = GetInputsWithoutInitializers(*model_proto);
-    input_features_ = builder.CreateDescriptorsFromValueInfoProtos(inputs);
-
-    // Create outputs
-    auto outputs = GetOutputs(*model_proto);
-    output_features_ = builder.CreateDescriptorsFromValueInfoProtos(outputs);
-
-    // author
-    auto has_producer_name = model_proto->has_producer_name();
-    author_ = has_producer_name
-                  ? model_proto->producer_name()
-                  : "";
-
-    // domain
-    auto has_domain = model_proto->has_domain();
-    domain_ = has_domain
-                  ? model_proto->domain()
-                  : "";
-
-    // name
-    auto has_graph = model_proto->has_graph();
-    auto graph_has_name = model_proto->graph().has_name();
-    auto is_name_available = has_graph && graph_has_name;
-    name_ = is_name_available
-                ? model_proto->graph().name()
-                : "";
-
-    // description
-    auto has_description = model_proto->has_doc_string();
-    description_ = has_description
-                       ? model_proto->doc_string()
-                       : "";
-
-    // version
-    auto has_version = model_proto->has_model_version();
-    version_ = has_version
-                   ? model_proto->model_version()
-                   : 0;
-  }
-};  // class ModelInfo
-
-class WinMLAdapter : public Microsoft::WRL::RuntimeClass<
-                         Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>,
-                         IWinMLAdapter> {
- private:
-  // TODO: Making this static is only temporary. A fix addressing the resulting the memory leaks is needed.
-  static std::shared_ptr<WinML::LotusEnvironment> lotus_environment_;
-
- public:
-  WinMLAdapter() {
-    if (lotus_environment_ == nullptr) {
-      lotus_environment_ = PheonixSingleton<WinML::LotusEnvironment>();
-    }
-  }
-  // factory methods for creating an ort model from a path
-  HRESULT STDMETHODCALLTYPE CreateModelProto(
-      const char* path,
-      IModelProto** model_proto) override try {
-    int file_descriptor;
-    _set_errno(0);  // clear errno
-    _sopen_s(
-        &file_descriptor,
-        path,
-        O_RDONLY | _O_SEQUENTIAL | _O_BINARY,
-        _SH_DENYWR,
-        _S_IREAD | _S_IWRITE);
-
-    errno_t err = 0;
-    _get_errno(&err);
-    THROW_HR_IF_MSG(
-        __HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND),
-        err == ENOENT,
-        "File not found: %s",
-        path);
-
-    THROW_HR_IF_MSG(
-        E_FAIL,
-        0 > file_descriptor,
-        "Failed");  //errno
-
-    auto stream = google::protobuf::io::FileInputStream(file_descriptor);
-    stream.SetCloseOnDelete(true);
-
-    auto model_proto_inner = new onnx::ModelProto();
-    THROW_HR_IF_MSG(
-        E_INVALIDARG,
-        model_proto_inner->ParseFromZeroCopyStream(&stream) == false,
-        "The stream failed to parse.");
-
-    auto model_proto_outer = wil::MakeOrThrow<ModelProto>(model_proto_inner);
-    return model_proto_outer.CopyTo(__uuidof(IModelProto), reinterpret_cast<void**>(model_proto));
-  }
-  WINMLA_CATCH_ALL_COM
-
-  // factory methods for creating an ort model from a stream
-  HRESULT STDMETHODCALLTYPE CreateModelProto(
-      ABI::Windows::Storage::Streams::IRandomAccessStreamReference* stream_reference,
-      IModelProto** model_proto) override try {
-    ZeroCopyInputStreamWrapper wrapper(stream_reference);
-
-    auto model_proto_inner = std::make_unique<onnx::ModelProto>();
-    THROW_HR_IF_MSG(
-        E_INVALIDARG,
-        model_proto_inner->ParseFromZeroCopyStream(&wrapper) == false,
-        "The stream failed to parse.");
-
-    auto model_proto_outer = wil::MakeOrThrow<ModelProto>(model_proto_inner.release());
-    return model_proto_outer.CopyTo(__uuidof(IModelProto), reinterpret_cast<void**>(model_proto));
-  }
-  WINMLA_CATCH_ALL_COM
-
-  // factory methods for creating an ort model from a model_proto
-  HRESULT STDMETHODCALLTYPE CreateModelProto(IModelProto* model_proto_in, IModelProto** model_proto) override try {
-    auto model_proto_inner = new onnx::ModelProto(*model_proto_in->get());
-    auto model_proto_outer = wil::MakeOrThrow<ModelProto>(model_proto_inner);
-    return model_proto_outer.CopyTo(__uuidof(IModelProto), reinterpret_cast<void**>(model_proto));
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE CreateModelInfo(IModelProto* model_proto, IModelInfo** model_info) override try {
-    auto model_info_outer = wil::MakeOrThrow<ModelInfo>(model_proto->get());
-    return model_info_outer.CopyTo(__uuidof(IModelInfo), reinterpret_cast<void**>(model_info));
-  }
-  WINMLA_CATCH_ALL_COM
-
-  void STDMETHODCALLTYPE EnableDebugOutput() override try {
-    WinML::CWinMLLogSink::EnableDebugOutput();
-  }
-  WINMLA_CATCH_ALL_DONOTHING
-
-  static bool IsFeatureDescriptorFp16(
-      winml::ILearningModelFeatureDescriptor descriptor) {
-    if (auto imageFeatureDescriptor = descriptor.try_as<winml::IImageFeatureDescriptor2>()) {
-      return TensorKind::Float16 == imageFeatureDescriptor.TensorKind();
-    }
-
-    if (auto tensorFeatureDescriptor = descriptor.try_as<winml::ITensorFeatureDescriptor>()) {
-      return TensorKind::Float16 == tensorFeatureDescriptor.TensorKind();
-    }
-
-    return false;
-  }
-
-  HRESULT STDMETHODCALLTYPE EnsureModelDeviceCompatibility(
-      winml::LearningModel const& model,
-      IModelProto* p_model_proto,
-      bool is_float16_supported) override try {
-    if (!is_float16_supported) {
-      auto& graph = p_model_proto->get()->graph();
-
-      // The model will not contain fp16 operations if:
-      // 1. The model has no fp16 inputs
-      // 2. The model has no fp16 initializers
-      // 3. The model does not create any fp16 intermediary tensors via the Cast (to float16) operator
-      // 4. The model does not have any fp16 outputs
-
-      // 1. Ensure that The model has no fp16 inputs
-      for (auto descriptor : model.InputFeatures()) {
-        THROW_HR_IF_MSG(
-            DXGI_ERROR_UNSUPPORTED,
-            IsFeatureDescriptorFp16(descriptor),
-            "The model contains a 16-bit input (%ls), but the current device does not support 16-bit float.",
-            descriptor.Name().c_str());
-      }
-
-      // 2. Ensure that the model has no fp16 initializers
-      for (int i = 0; i < graph.node_size(); i++) {
-        auto node = graph.node(i);
-        if (node.op_type() == "Cast" && node.domain().empty()) {
-          for (int attribIndex = 0; attribIndex < node.attribute_size(); attribIndex++) {
-            auto attribute = node.attribute(attribIndex);
-            if (attribute.name() == "to") {
-              THROW_HR_IF_MSG(
-                  DXGI_ERROR_UNSUPPORTED,
-                  attribute.i() == onnx::TensorProto::DataType::TensorProto_DataType_FLOAT16,
-                  "The model contains a 16-bit float Cast Op (%s), but the current device does not support 16-bit float.",
-                  node.name().c_str());
-            }
-          }
-        }
-      }
-
-      // 3. Ensure that the model does not create any fp16 intermediary
-      //    tensors via the Cast (to float16) operator
-      for (int i = 0; i < graph.initializer_size(); i++) {
-        auto initializer = graph.initializer(i);
-
-        THROW_HR_IF_MSG(
-            DXGI_ERROR_UNSUPPORTED,
-            initializer.data_type() == onnx::TensorProto::DataType::TensorProto_DataType_FLOAT16,
-            "The model contains a 16-bit float initializer (%s), but the current device does not support 16-bit float.",
-            initializer.name().c_str());
-      }
-
-      // 4. Ensure that the model does not have any fp16 outputs
-      for (auto descriptor : model.OutputFeatures()) {
-        THROW_HR_IF_MSG(
-            DXGI_ERROR_UNSUPPORTED,
-            IsFeatureDescriptorFp16(descriptor),
-            "The model contains a 16-bit output (%ls), but the current device does not support 16-bit float.",
-            descriptor.Name().c_str());
-      }
-    }
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  ID3D12Resource* STDMETHODCALLTYPE GetD3D12ResourceFromAllocation(onnxruntime::IExecutionProvider* provider, void* allocation) override try {
-#ifdef USE_DML
-    auto d3dResource =
-        Dml::GetD3D12ResourceFromAllocation(
-            provider->GetAllocator(0, ::OrtMemType::OrtMemTypeDefault).get(),
-            allocation);
-    return d3dResource;
-#else
-    return nullptr;
-#endif USE_DML
-  } catch (...) {
-    return nullptr;
-  }
-
-  static onnxruntime::MLDataType GetType(winml::TensorKind kind) {
-    switch (kind) {
-      case winml::TensorKind::Float:
-        return onnxruntime::DataTypeImpl::GetType<float>();
-      case winml::TensorKind::Float16:
-        return onnxruntime::DataTypeImpl::GetType<onnxruntime::MLFloat16>();
-    };
-    return nullptr;
-  }
-
-  // factory method for creating an ortsessionbuilder from a device
-  HRESULT STDMETHODCALLTYPE CreateOrtSessionBuilder(
-      ID3D12Device* device,
-      ID3D12CommandQueue* queue,
-      IOrtSessionBuilder** session_builder) override try {
-    if (device == nullptr) {
-      auto builder = wil::MakeOrThrow<CpuOrtSessionBuilder>();
-      return builder.CopyTo(__uuidof(IOrtSessionBuilder), reinterpret_cast<void**>(session_builder));
-    }
-#ifdef USE_DML
-    else {
-      auto builder = wil::MakeOrThrow<DmlOrtSessionBuilder>(device, queue);
-      return builder.CopyTo(__uuidof(IOrtSessionBuilder), reinterpret_cast<void**>(session_builder));
-    }
-#else
-    return E_NOTIMPL;
-#endif USE_DML
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetMapType(const OrtValue* ort_value, ONNXTensorElementDataType* key_type, ONNXTensorElementDataType* value_type) override try {
-    *key_type = *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_UNDEFINED;
-    auto type = ort_value->Type();
-    if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapStringToString>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapStringToInt64>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapStringToFloat>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapStringToDouble>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_DOUBLE;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapInt64ToString>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapInt64ToInt64>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapInt64ToFloat>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::MapInt64ToDouble>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_DOUBLE;
-    }
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetVectorMapType(const OrtValue* ort_value, ONNXTensorElementDataType* key_type, ONNXTensorElementDataType* value_type) override try {
-    *key_type = *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_UNDEFINED;
-    auto type = ort_value->Type();
-    if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::VectorMapStringToFloat>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT;
-    } else if (type == onnxruntime::DataTypeImpl::GetType<onnxruntime::VectorMapInt64ToFloat>()) {
-      *key_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64;
-      *value_type = ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT;
-    }
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetCustomRegistry(IMLOperatorRegistry** registry) override try {
-#ifdef USE_DML
-    auto impl = wil::MakeOrThrow<AbiCustomRegistryImpl>();
-    *registry = impl.Detach();
-    return S_OK;
-#else
-    return E_NOTIMPL;
-#endif USE_DML
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetOperatorRegistry(ILearningModelOperatorProviderNative* operator_provider_native, IMLOperatorRegistry** registry) override try {
-#ifdef USE_DML
-    // Retrieve the "operator abi" registry.
-    winrt::com_ptr<IMLOperatorRegistry> operator_registry;
-    THROW_IF_FAILED(operator_provider_native->GetRegistry(operator_registry.put()));
-    *registry = operator_registry.detach();
-    return S_OK;
-#else
-    return E_NOTIMPL;
-#endif USE_DML
-  }
-  WINMLA_CATCH_ALL_COM
-
-  void* STDMETHODCALLTYPE CreateGPUAllocationFromD3DResource(ID3D12Resource* pResource) override try {
-#ifdef USE_DML
-    return Dml::CreateGPUAllocationFromD3DResource(pResource);
-#else
-    return nullptr;
-#endif USE_DML
-  } catch (...) {
-    return nullptr;
-  }
-
-  void STDMETHODCALLTYPE FreeGPUAllocation(void* ptr) override try {
-#ifdef USE_DML
-    Dml::FreeGPUAllocation(ptr);
-#endif USE_DML
-  }
-  WINMLA_CATCH_ALL_DONOTHING
-
-  HRESULT STDMETHODCALLTYPE CopyTensor(
-      onnxruntime::IExecutionProvider* provider,
-      OrtValue* src,
-      OrtValue* dst) override try {
-#ifdef USE_DML
-    ORT_THROW_IF_ERROR(Dml::CopyTensor(provider, *(src->GetMutable<onnxruntime::Tensor>()), *(dst->GetMutable<onnxruntime::Tensor>())));
-    return S_OK;
-#else
-    return E_NOTIMPL;
-#endif USE_DML
-  }
-  WINMLA_CATCH_ALL_COM
-
-  // Override select shape inference functions which are incomplete in ONNX with versions that are complete,
-  // and are also used in DML kernel registrations.  Doing this avoids kernel and shader creation being
-  // deferred until first evaluation.  It also prevents a situation where inference functions in externally
-  // registered schema are reachable only after upstream schema have been revised in a later OS release,
-  // which would be a compatibility risk.
-  HRESULT STDMETHODCALLTYPE OverrideSchemaInferenceFunctions() override try {
-#ifdef USE_DML
-    static std::once_flag schema_override_once_flag;
-    std::call_once(schema_override_once_flag, []() {
-      SchemaInferenceOverrider::OverrideSchemaInferenceFunctions();
-    });
-    return S_OK;
-#else
-    return S_OK;  // needs to return S_OK otherwise everything breaks because this gets called from the learningmodel constructor
-#endif USE_DML
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetProviderMemoryInfo(
-      onnxruntime::IExecutionProvider* provider,
-      OrtMemoryInfo** memory_info) override try {
-    auto allocator = provider->GetAllocator(0, ::OrtMemType::OrtMemTypeDefault);
-
-    const auto& info = allocator->Info();
-    *memory_info = new OrtMemoryInfo(info.name, info.alloc_type, info.device, info.id, info.mem_type);
-    if (*memory_info == nullptr) {
-      return E_OUTOFMEMORY;
-    }
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE GetValueMemoryInfo(const OrtValue* ort_value, OrtMemoryInfo** memory_info) override try {
-    const auto& tensor = ort_value->Get<onnxruntime::Tensor>();
-    auto info = tensor.Location();
-    *memory_info = new OrtMemoryInfo(info.name, info.alloc_type, info.device, info.id, info.mem_type);
-    if (*memory_info == nullptr) {
-      return E_OUTOFMEMORY;
-    }
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  struct AllocatorWrapper : public OrtAllocator {
-   public:
-    AllocatorWrapper(onnxruntime::AllocatorPtr impl) : impl_(impl) {
-      version = ORT_API_VERSION;
-      Alloc = AllocImpl;
-      Free = FreeImpl;
-      Info = InfoImpl;
-    }
-
-    static void* ORT_API_CALL AllocImpl(struct OrtAllocator* this_, size_t size) {
-      return static_cast<AllocatorWrapper*>(this_)->impl_->Alloc(size);
-    }
-    static void ORT_API_CALL FreeImpl(struct OrtAllocator* this_, void* p) {
-      return static_cast<AllocatorWrapper*>(this_)->impl_->Free(p);
-    }
-    static const struct OrtMemoryInfo* ORT_API_CALL InfoImpl(const struct OrtAllocator* this_) {
-      return &(static_cast<const AllocatorWrapper*>(this_)->impl_->Info());
-    }
-
-   private:
-    onnxruntime::AllocatorPtr impl_;
-  };
-
-  HRESULT STDMETHODCALLTYPE GetProviderAllocator(
-      onnxruntime::IExecutionProvider* provider,
-      OrtAllocator** allocator) override try {
-    auto allocator_ptr = provider->GetAllocator(0, ::OrtMemType::OrtMemTypeDefault);
-    *allocator = new (std::nothrow) AllocatorWrapper(allocator_ptr);
-    if (*allocator == nullptr) {
-      return E_OUTOFMEMORY;
-    }
-
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-
-  HRESULT STDMETHODCALLTYPE FreeProviderAllocator(
-      OrtAllocator* allocator) override try {
-    delete static_cast<AllocatorWrapper*>(allocator);
-    return S_OK;
-  }
-  WINMLA_CATCH_ALL_COM
-};  // namespace Windows::AI::MachineLearning::Adapter
-std::shared_ptr<WinML::LotusEnvironment> WinMLAdapter::lotus_environment_ = nullptr;
-
-extern "C" HRESULT STDMETHODCALLTYPE OrtGetWinMLAdapter(IWinMLAdapter** adapter) try {
-  // make an adapter instance
-  Microsoft::WRL::ComPtr<WinMLAdapter> adapterptr = wil::MakeOrThrow<WinMLAdapter>();
-  return adapterptr.CopyTo(__uuidof(IWinMLAdapter), reinterpret_cast<void**>(adapter));
-}
-WINMLA_CATCH_ALL_COM
-
-// InferenceSession
-// ================
-
-InferenceSession::InferenceSession(onnxruntime::InferenceSession* session) : session_(session) {
-}
-
-void STDMETHODCALLTYPE InferenceSession::RegisterGraphTransformers() try {
-#ifdef USE_DML
-  // Bug 22973884 : Fix issues with BatchNorm + Add and BatchNorm + Mul handling implicit inputs, and move from Winml to ORT
-  GraphTransformerHelpers::RegisterGraphTransformers(session_.get());
-#endif USE_DML
-}
-WINMLA_CATCH_ALL_DONOTHING
-
-HRESULT STDMETHODCALLTYPE InferenceSession::StartProfiling() try {
-  this->session_->StartProfiling(PheonixSingleton<WinML::LotusEnvironment>()->GetDefaultLogger());
-  return S_OK;
-}
-WINMLA_CATCH_ALL_COM
-
-HRESULT STDMETHODCALLTYPE InferenceSession::EndProfiling() try {
-  this->session_->EndProfiling();
-  return S_OK;
-}
-WINMLA_CATCH_ALL_COM
-
-HRESULT STDMETHODCALLTYPE
-InferenceSession::LoadModel(
-    IModelProto* model_proto) try {
-  auto session_protected_load_accessor =
-      static_cast<InferenceSessionProtectedLoadAccessor*>(session_.get());
-  // session's like to have their very own copy of the model_proto, use detach()
-  std::unique_ptr<ONNX_NAMESPACE::ModelProto> model_proto_ptr(model_proto->detach());
-  ORT_THROW_IF_ERROR(session_protected_load_accessor->Load(std::move(model_proto_ptr)));
-  return S_OK;
-}
-WINMLA_CATCH_ALL_COM
-
-HRESULT STDMETHODCALLTYPE
-InferenceSession::RegisterCustomRegistry(
-    IMLOperatorRegistry* registry) try {
-  RETURN_HR_IF(S_OK, registry == nullptr);
-
-#ifdef USE_DML
-  auto custom_registries = GetLotusCustomRegistries(registry);
-
-  // Register
-  for (auto& custom_registry : custom_registries) {
-    ORT_THROW_IF_ERROR(session_->RegisterCustomRegistry(custom_registry));
-  }
-#endif USE_DML
-
-  return S_OK;
-}
-WINMLA_CATCH_ALL_COM
-
-void STDMETHODCALLTYPE InferenceSession::FlushContext(onnxruntime::IExecutionProvider* dml_provider) try {
-#ifdef USE_DML
-  Dml::FlushContext(dml_provider);
-#endif USE_DML
-}
-WINMLA_CATCH_ALL_DONOTHING
-
-void STDMETHODCALLTYPE InferenceSession::TrimUploadHeap(onnxruntime::IExecutionProvider* dml_provider) try {
-#ifdef USE_DML
-  Dml::TrimUploadHeap(dml_provider);
-#endif USE_DML
-}
-WINMLA_CATCH_ALL_DONOTHING
-
-void STDMETHODCALLTYPE InferenceSession::ReleaseCompletedReferences(onnxruntime::IExecutionProvider* dml_provider) try {
-#ifdef USE_DML
-  Dml::ReleaseCompletedReferences(dml_provider);
-#endif USE_DML
-}
-WINMLA_CATCH_ALL_DONOTHING
-
-HRESULT STDMETHODCALLTYPE InferenceSession::CopyOneInputAcrossDevices(
-    const char* input_name,
-    const OrtValue* orig_mlvalue,
-    OrtValue** new_mlvalue) try {
-  auto session_protected_load_accessor =
-      static_cast<InferenceSessionProtectedLoadAccessor*>(session_.get());
-  const onnxruntime::SessionState& sessionState = session_protected_load_accessor->GetSessionState();
-  auto temp_mlvalue = std::make_unique<OrtValue>();
-  ORT_THROW_IF_ERROR(onnxruntime::utils::CopyOneInputAcrossDevices(sessionState, input_name, *orig_mlvalue, *temp_mlvalue.get()));
-  *new_mlvalue = temp_mlvalue.release();
-  return S_OK;
-}
-WINMLA_CATCH_ALL_COM
-
-}  // namespace Windows::AI::MachineLearning::Adapter
\ No newline at end of file