[tmva][pymva] Use .h5 file instead of .keras

Go back to use .h5 files instead of .keras due to a problem in Keras loading .keras files on MacOS ARM (see https://github.com/keras-team/keras/issues/18278 )

math/mathcore/src/FitUtil.cxx

@@ -1118,7 +1118,7 @@ double FitUtil::EvaluateLogL(const IModelFunction &func, const UnBinData &data,
          } else {
             // use (-inf +inf)
             data.Range().GetRange(&xmin[0],&xmax[0]);
-            // check if funcition is zero at +- inf
+            // check if function is zero at +- inf
             if (func(xmin.data(), p) != 0 || func(xmax.data(), p) != 0) {
                MATH_ERROR_MSG("FitUtil::EvaluateLogLikelihood","A range has not been set and the function is not zero at +/- inf");
                return 0;

tmva/pymva/test/EmitCustomModel.cxx

@@ -34,7 +34,7 @@ y_train=randomGenerator.rand(1,4)\n\
 \n\
 model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))\n\
 model.fit(x_train, y_train, epochs=10, batch_size=1)\n\
-model.save('KerasModelForCustomOp.keras')\n";
+model.save('KerasModelForCustomOp.h5')\n";
 
 int main(){
    Py_Initialize();
@@ -43,7 +43,7 @@ int main(){
    PyRun_SimpleString(pythonSrc);
 
     //Parsing the saved Keras .keras file into RModel object
-    RModel model = PyKeras::Parse("KerasModelForCustomOp.keras");
+    RModel model = PyKeras::Parse("KerasModelForCustomOp.h5");
     model.Generate();
 
     std::unique_ptr<ROperator> op;

tmva/pymva/test/EmitFromKeras.cxx

@@ -21,54 +21,54 @@ int main(){
    PyRun_SimpleFile(fKerasModels, "generateKerasModels.py");
 
    //Emitting header file for Keras Sequential Model
-   RModel modelSequential = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelSequential.keras");
+   RModel modelSequential = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelSequential.h5");
    modelSequential.Generate();
    modelSequential.OutputGenerated("KerasSequentialModel.hxx");
 
    //Emitting header file for Keras Functional API Model
-   RModel modelFunctional = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelFunctional.keras");
+   RModel modelFunctional = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelFunctional.h5");
    modelFunctional.Generate();
    modelFunctional.OutputGenerated("KerasFunctionalModel.hxx");
 
    //Emitting header file for Keras BatchNorm Model
-   RModel modelBatchNorm = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelBatchNorm.keras");
+   RModel modelBatchNorm = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelBatchNorm.h5");
    modelBatchNorm.Generate();
    modelBatchNorm.OutputGenerated("KerasBatchNormModel.hxx");
 
 #if PY_MAJOR_VERSION >= 3  // parsing of convolutional models supported only for Python3
    // Emitting header file for Keras Conv2D Model with valid padding
-   RModel modelConv2D_Valid = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelConv2D_Valid.keras");
+   RModel modelConv2D_Valid = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelConv2D_Valid.h5");
    modelConv2D_Valid.Generate();
    modelConv2D_Valid.OutputGenerated("KerasConv2D_Valid.hxx");
 
    // Emitting header file for Keras Conv2D Model with valid padding
-   RModel modelConv2D_Same = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelConv2D_Same.keras");
+   RModel modelConv2D_Same = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelConv2D_Same.h5");
    modelConv2D_Same.Generate();
    modelConv2D_Same.OutputGenerated("KerasConv2D_Same.hxx");
 #endif
 
    //Emitting header file for Keras model with Reshape layer
-   RModel modelReshape = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelReshape.keras");
+   RModel modelReshape = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelReshape.h5");
    modelReshape.Generate();
    modelReshape.OutputGenerated("KerasReshapeModel.hxx");
 
    //Emitting header file for Keras model with Concatenate layer
-   RModel modelConcatenate = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelConcatenate.keras");
+   RModel modelConcatenate = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelConcatenate.h5");
    modelConcatenate.Generate();
    modelConcatenate.OutputGenerated("KerasConcatenateModel.hxx");
 
    // Emitting header file for Keras Binary Op Model
-   RModel modelBinaryOp = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelBinaryOp.keras");
+   RModel modelBinaryOp = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelBinaryOp.h5");
    modelBinaryOp.Generate();
    modelBinaryOp.OutputGenerated("KerasBinaryOpModel.hxx");
 
    //Emitting header file for Keras activation functions model
-   RModel modelActivations = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelActivations.keras");
+   RModel modelActivations = TMVA::Experimental::SOFIE::PyKeras::Parse("KerasModelActivations.h5");
    modelActivations.Generate();
    modelActivations.OutputGenerated("KerasActivationsModel.hxx");
 
    // Emitting header file for Swish activation functions model
-   RModel modelSwish = TMVA::Experimental::SOFIE::PyKeras::Parse("swish_model.keras");
+   RModel modelSwish = TMVA::Experimental::SOFIE::PyKeras::Parse("swish_model.h5");
    modelSwish.Generate();
    modelSwish.OutputGenerated("KerasSwish_model.hxx");
 

tmva/pymva/test/generateKerasModels.py

@@ -24,7 +24,7 @@ def generateFunctionalModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=2)
-    model.save('KerasModelFunctional.keras')
+    model.save('KerasModelFunctional.h5')
 
 def generateSequentialModel():
     model=Sequential()
@@ -39,7 +39,7 @@ def generateSequentialModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=4)
-    model.save('KerasModelSequential.keras')
+    model.save('KerasModelSequential.h5')
 
 def generateBatchNormModel():
     model=Sequential()
@@ -53,7 +53,7 @@ def generateBatchNormModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=2)
-    model.save('KerasModelBatchNorm.keras')
+    model.save('KerasModelBatchNorm.h5')
 
 def generateConv2DModel_ValidPadding():
     model=Sequential()
@@ -65,7 +65,7 @@ def generateConv2DModel_ValidPadding():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=2)
-    model.save('KerasModelConv2D_Valid.keras')
+    model.save('KerasModelConv2D_Valid.h5')
 
 def generateConv2DModel_SamePadding():
     model=Sequential()
@@ -77,7 +77,7 @@ def generateConv2DModel_SamePadding():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=2)
-    model.save('KerasModelConv2D_Same.keras')
+    model.save('KerasModelConv2D_Same.h5')
 
 def generateReshapeModel():
     model = Sequential()
@@ -90,7 +90,7 @@ def generateReshapeModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=2)
-    model.save('KerasModelReshape.keras')
+    model.save('KerasModelReshape.h5')
 
 def generateConcatModel():
     input_1 = Input(shape=(2,))
@@ -107,7 +107,7 @@ def generateConcatModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit([x1_train,x2_train], y_train, epochs=10, batch_size=1)
-    model.save('KerasModelConcatenate.keras')
+    model.save('KerasModelConcatenate.h5')
 
 def generateBinaryOpModel():
     input1 = Input(shape=(2, ))
@@ -124,7 +124,7 @@ def generateBinaryOpModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit([x1_train,x2_train], y_train, epochs=10, batch_size=2)
-    model.save('KerasModelBinaryOp.keras')
+    model.save('KerasModelBinaryOp.h5')
 
 def generateActivationModel():
     input=Input(shape=(8,))
@@ -140,7 +140,7 @@ def generateActivationModel():
 
     model.compile(loss='mean_squared_error', optimizer=SGD(learning_rate=0.01))
     model.fit(x_train, y_train, epochs=10, batch_size=1)
-    model.save('KerasModelActivations.keras')
+    model.save('KerasModelActivations.h5')
 
 def generateSwishModel():
 # Create the Keras model
@@ -153,7 +153,7 @@ def generateSwishModel():
               optimizer='adam',
               metrics=['accuracy'])
     # Save the model as an.keras file
-    model.save('swish_model.keras')
+    model.save('swish_model.h5')
 
 
 generateFunctionalModel()

tmva/pymva/test/testPyKerasClassification.C

@@ -30,19 +30,20 @@ int testPyKerasClassification(){
    TFile *input = TFile::Open(fname);
 
    // Build model from python file
-   std::cout << "Generate keras model..." << std::endl;
-   UInt_t ret;
-   ret = gSystem->Exec("echo '"+pythonSrc+"' > generateKerasModelClassification.py");
-   if(ret!=0){
-       std::cout << "[ERROR] Failed to write python code to file" << std::endl;
-       return 1;
+   if (gSystem->AccessPathName("kerasModelClassification.h5")) {
+      std::cout << "Generate keras model..." << std::endl;
+      UInt_t ret;
+      ret = gSystem->Exec("echo '"+pythonSrc+"' > generateKerasModelClassification.py");
+      if(ret!=0){
+         std::cout << "[ERROR] Failed to write python code to file" << std::endl;
+         return 1;
+      }
+      ret = gSystem->Exec(TMVA::Python_Executable() + " generateKerasModelClassification.py");
+      if(ret!=0){
+         std::cout << "[ERROR] Failed to generate model using python" << std::endl;
+         return 1;
+      }
    }
-   ret = gSystem->Exec(TMVA::Python_Executable() + " generateKerasModelClassification.py");
-   if(ret!=0){
-       std::cout << "[ERROR] Failed to generate model using python" << std::endl;
-       return 1;
-   }
-
    // Setup PyMVA and factory
    std::cout << "Setup TMVA..." << std::endl;
    TMVA::PyMethodBase::PyInitialize();