Add functions to allow ILearningModelFeatureValues to be plumbed through as input through WinMLRunner static library (#327)

* Make changes to allow protobuf input through commandline args

* use protobuf helper

* Use feature values instead of protobuf

* Commandline args changes

* added clear method

* Naming changes

* Formatting

* DOn't modify git modules

Tools/WinMLRunner/onnxruntime

@@ -0,0 +1 @@
+Subproject commit ccbf49e59f6bef897a94595e2213263b37d64ff3

Tools/WinMLRunner/src/CommandLineArgs.cpp

@@ -600,3 +600,7 @@ void CommandLineArgs::AddPerformanceFileMetadata(const std::string& key, const s
     cleanedValue.erase(std::remove_copy(value.begin(), value.end(), cleanedValue.begin(), ','), cleanedValue.end());
     m_perfFileMetadata.push_back(std::make_pair(key, cleanedValue));
 }
+void CommandLineArgs::AddProvidedInputFeatureValue(const ILearningModelFeatureValue& input)
+{
+    m_providedInputFeatureValues.push_back(input);
+}

Tools/WinMLRunner/src/CommandLineArgs.h

@@ -45,6 +45,10 @@ public:
     BitmapInterpolationMode AutoScaleInterpMode() const { return m_autoScaleInterpMode; }
 
     const std::vector<std::wstring>& ImagePaths() const { return m_imagePaths; }
+    const std::vector<ILearningModelFeatureValue>& ProvidedInputFeatureValues() const
+    {
+        return m_providedInputFeatureValues;
+    }
     const std::wstring& CsvPath() const { return m_csvData; }
     const std::wstring& OutputPath() const { return m_perfOutputPath; }
     const std::wstring& FolderPath() const { return m_modelFolderPath; }
@@ -92,11 +96,11 @@ public:
     bool IsGarbageInput() const
     {
         // When there is no image or csv input provided, then garbage input binding is used.
-        return m_imagePaths.empty() && m_csvData.empty();
+        return m_imagePaths.empty() && m_csvData.empty() && m_providedInputFeatureValues.empty();
     }
     bool IsCSVInput() const { return m_imagePaths.empty() && !m_csvData.empty(); }
     bool IsImageInput() const { return !m_imagePaths.empty() && m_csvData.empty(); }
-
+    bool InputFeatureValuesProvided() const { return !m_providedInputFeatureValues.empty(); }
     uint32_t NumIterations() const { return m_numIterations; }
     uint32_t NumLoadIterations() const { return m_numLoadIterations; }
     uint32_t NumSessionCreationIterations() const { return m_numSessionIterations; }
@@ -140,6 +144,8 @@ public:
     void SetSessionCreationIterations(const uint32_t iterations) { m_numSessionIterations = iterations; }
     void SetLoadIterations(const uint32_t iterations) { m_numLoadIterations = iterations; }
     void AddPerformanceFileMetadata(const std::string& key, const std::string& value);
+    void AddProvidedInputFeatureValue(const ILearningModelFeatureValue& input);
+    void ClearProvidedInputFeatureValues() { m_providedInputFeatureValues.clear(); };
     void SetGarbageDataMaxValue(const uint32_t value) { m_garbageDataMaxValue = value; }
 
     // Stop iterating when total time of iterations after the first iteration exceeds time limit.
@@ -185,6 +191,7 @@ private:
     std::wstring m_modelFolderPath;
     std::wstring m_modelPath;
     std::vector<std::wstring> m_imagePaths;
+    std::vector<ILearningModelFeatureValue> m_providedInputFeatureValues;
     std::wstring m_inputImageFolderPath;
     std::wstring m_csvData;
     std::wstring m_inputData;

Tools/WinMLRunner/src/Run.cpp

@@ -221,17 +221,23 @@ HRESULT BindInputs(LearningModelBinding& context, const LearningModelSession& se
     bool captureIterationPerf = args.IsPerformanceCapture() || args.IsPerIterationCapture();
 
     std::vector<ILearningModelFeatureValue> inputFeatures;
-    try
+    if (args.InputFeatureValuesProvided())
     {
-        inputFeatures = GenerateInputFeatures(session.Model(), args, inputBindingType, inputDataType, device, iteration, imagePath);
+        inputFeatures = args.ProvidedInputFeatureValues();
     }
-    catch (hresult_error hr)
+    else
     {
-        std::wcout << "\nGenerating Input Features [FAILED]" << std::endl;
-        std::wcout << hr.message().c_str() << std::endl;
-        return hr.code();
+        try
+        {
+            inputFeatures = GenerateInputFeatures(session.Model(), args, inputBindingType, inputDataType, device, iteration, imagePath);
+        }
+        catch (hresult_error hr)
+        {
+            std::wcout << "\nGenerating Input Features [FAILED]" << std::endl;
+            std::wcout << hr.message().c_str() << std::endl;
+            return hr.code();
+        }
     }
-
     HRESULT bindInputResult =
         BindInputFeatures(session.Model(), context, inputFeatures, args, output, captureIterationPerf, iteration, profiler);