[tmva][pymva] Fix and improve Python tutorials and set tf.keras=1 by default

- protect usage of pymva when it is available
- set tf.keras by default in MethodPyKeras since this is now the standard for
Keras.

Author: lmoneta

tmva/pymva/inc/TMVA/MethodPyKeras.h

@@ -86,7 +86,7 @@ namespace TMVA {
       UInt_t fNumEpochs {0}; // Number of training epochs
       Int_t fNumThreads {0}; // Number of CPU threads (if 0 uses default values)
       Int_t fVerbose; // Keras verbosity during training
-      Bool_t fUseTFKeras { kFALSE};   // use Keras from Tensorflow (-1, default, 0 false, 1, true)
+      Bool_t fUseTFKeras { true};   // use Keras from Tensorflow default is true
       Bool_t fContinueTraining; // Load weights from previous training
       Bool_t fSaveBestOnly; // Store only weights with smallest validation loss
       Int_t fTriesEarlyStopping; // Stop training if validation loss is not decreasing for several epochs

tmva/pymva/src/MethodPyKeras.cxx

@@ -82,6 +82,7 @@ void MethodPyKeras::DeclareOptions() {
    DeclareOptionRef(fNumThreads, "NumThreads", "Number of CPU threads (only for Tensorflow backend)");
    DeclareOptionRef(fGpuOptions, "GpuOptions", "GPU options for tensorflow, such as allow_growth");
    DeclareOptionRef(fUseTFKeras, "tf.keras", "Use tensorflow from Keras");
+   DeclareOptionRef(fUseTFKeras, "tfkeras", "Use tensorflow from Keras");
    DeclareOptionRef(fVerbose, "Verbose", "Keras verbosity during training");
    DeclareOptionRef(fContinueTraining, "ContinueTraining", "Load weights from previous training");
    DeclareOptionRef(fSaveBestOnly, "SaveBestOnly", "Store only weights with smallest validation loss");

tutorials/tmva/TMVA_CNN_Classification.py

@@ -33,8 +33,6 @@
 import importlib
 
 TMVA.Tools.Instance()
-TMVA.PyMethodBase.PyInitialize()
-
 
 def MakeImagesTree(n, nh, nw):
     # image size (nh x nw)
@@ -122,6 +120,8 @@ def MakeImagesTree(n, nh, nw):
 if ROOT.gSystem.GetFromPipe("root-config --has-tmva-pymva") != "yes":
     useKerasCNN = False
     usePyTorchCNN = False
+else:
+    TMVA.PyMethodBase.PyInitialize()
 
 tf_spec = importlib.util.find_spec("tensorflow")
 if tf_spec is None:
@@ -139,10 +139,6 @@ def MakeImagesTree(n, nh, nw):
         "TMVA is not build with GPU or CPU multi-thread support. Cannot use TMVA Deep Learning for CNN",
     )
 
-#there is an issue using TF and PyTorch - so use only one of the two
-if (usePyTorchCNN):
-   useKerasCNN = False
-
 writeOutputFile = True
 
 num_threads = 0  # use default threads

tutorials/tmva/TMVA_Higgs_Classification.py

@@ -30,15 +30,12 @@
 ## - The third argument is a string option defining some general configuration for the TMVA session. For example all TMVA output can be suppressed by removing the "!" (not) in front of the "Silent" argument in the option string
 
 import ROOT
+import os
 
 TMVA = ROOT.TMVA
 TFile = ROOT.TFile
 
-
 TMVA.Tools.Instance()
-## For PYMVA methods
-TMVA.PyMethodBase.PyInitialize()
-
 
 # options to control used methods
 useLikelihood = True  # likelihood based discriminant
@@ -47,8 +44,20 @@
 useMLP = False  # Multi Layer Perceptron (old TMVA NN implementation)
 useBDT = True  # Boosted Decision Tree
 useDL = True  # TMVA Deep learning ( CPU or GPU)
+useKeras = True # Use Keras Deep Learning via PyMVA
+
+if ROOT.gSystem.GetFromPipe("root-config --has-tmva-pymva") == "yes":
+    TMVA.PyMethodBase.PyInitialize()
+else:
+    useKeras = False  # cannot use Keras if PYMVA is not available
+
+if useKeras:
+    try:
+        import tensorflow
+    except:
+        ROOT.Warning("TMVA_Higgs_Classification", "Skip using Keras since tensorflow is not available")
+        useKeras = False
 
-TMVA.Tools.Instance()
 outputFile = TFile.Open("Higgs_ClassificationOutput.root", "RECREATE")
 factory = TMVA.Factory(
     "TMVA_Higgs_Classification", outputFile, V=False, ROC=True, Silent=False, Color=True, AnalysisType="Classification"
@@ -276,7 +285,7 @@
     training1 = ROOT.TString(
         "LearningRate=1e-3,Momentum=0.9,"
         "ConvergenceSteps=10,BatchSize=128,TestRepetitions=1,"
-        "MaxEpochs=30,WeightDecay=1e-4,Regularization=None,"
+        "MaxEpochs=20,WeightDecay=1e-4,Regularization=None,"
         "Optimizer=ADAM,ADAM_beta1=0.9,ADAM_beta2=0.999,ADAM_eps=1.E-7,"  # ADAM default parameters
         "DropConfig=0.0+0.0+0.0+0."
     )
@@ -310,6 +319,45 @@
         Architecture=arch,
     )
 
+#Keras DL
+if useKeras:
+    ROOT.Info("TMVA_Higgs_Classification", "Building Deep Learning keras model")
+    # create Keras model with 4 layers of 64 units and relu activations
+    import tensorflow
+    from tensorflow.keras.models import Sequential
+    from tensorflow.keras.optimizers import Adam
+    from tensorflow.keras.layers import Input, Dense
+
+    model = Sequential()
+    model.add(Dense(64, activation='relu',input_dim=7))
+    model.add(Dense(64, activation='relu'))
+    model.add(Dense(64, activation='relu'))
+    model.add(Dense(64, activation='relu'))
+    model.add(Dense(2, activation='sigmoid'))
+    model.compile(loss = 'binary_crossentropy', optimizer = Adam(learning_rate = 0.001), metrics = ['accuracy'])
+    model.save('model_higgs.h5')
+    model.summary()
+
+    if not os.path.exists("model_higgs.h5"):
+        raise FileNotFoundError("Error creating Keras model file - skip using Keras")
+    else:
+        # book PyKeras method only if Keras model could be created
+        ROOT.Info("TMVA_Higgs_Classification", "Booking Deep Learning keras model")
+        factory.BookMethod(
+            loader,
+            TMVA.Types.kPyKeras,
+            "PyKeras",
+            H=True,
+            V=False,
+            VarTransform=None,
+            FilenameModel="model_higgs.h5",
+            FilenameTrainedModel="trained_model_higgs.h5",
+            NumEpochs=20,
+            BatchSize=100 )
+#            GpuOptions="allow_growth=True",
+#        )  # needed for RTX NVidia card and to avoid TF allocates all GPU memory
+
+
 ## Train Methods
 
 # Here we train all the previously booked methods.