Evaluating multiple TMVA (pyKeras) weights using the same intree variables?

TMVA
1

I have trained multiple CNN models using the TMVA Classification code. I can evaluate a single CNN model weight in my code just fine, and I’m happy with the results it’s producing. I’m now trying to add multiple weights, the goal is to use a combination of selection cuts from multiple CNN weights for my analysis. However, whenever I try to evaluate more than one weight, my code crashes. It uses the same variables in my input ROOT file TTree, so I suspect I’m doing something incorrectly. I tried using multiple outtrees and reader variables, but still no luck. Here’s a snippet of my current code:

//skipping header files 

#include "/home/software/root/root_src/tmva/tmva/inc/TMVA/Factory.h"
#include "/home/software/root/root_src/tmva/tmva/inc/TMVA/Tools.h"
#include "/home/software/root/root_src/tmva/tmva/inc/TMVA/Reader.h"
#include "/home/software/root/root_src/tmva/tmva/inc/TMVA/MethodCuts.h"
#include "/home/software/root/root_src/tmva/pymva/inc/TMVA/PyMethodBase.h"

#include <TSystem.h>

  //reading input file
  TChain* intree = new TChain("data");

  float elost00xlad00; intree->SetBranchAddress("elost00xlad00",&elost00xlad00);
  float elost00xlad01; intree->SetBranchAddress("elost00xlad01",&elost00xlad01);
...
  float elost19xlad16; intree->SetBranchAddress("elost19xlad16",&elost19xlad16);
  float elost19xlad17; intree->SetBranchAddress("elost19xlad17",&elost19xlad17);


  TFile *outfile = new TFile(Form("%s",outfname.Data()), "RECREATE", "", ROOT::CompressionSettings(ROOT::kLZMA, 8));
  TTree *outtree = new TTree("cleandata", "cleandata");

outtree->Branch("var1",&elost00xlad00);
outtree->Branch("var2",&elost00xlad01);
...
outtree->Branch("var359",&elost00xlad00);
outtree->Branch("var360",&elost19xlad17);

// PyKeras response value
float pykeras; outtree->Branch("pykeras",&pykeras); 
float pykeras2; outtree->Branch("pykeras2",&pykeras2); 

 // Setup TMVA for PyKeras
 
   TMVA::Tools::Instance();
   TMVA::PyMethodBase::PyInitialize();


// Create a TMVA reader object for pykeras CNN models
TMVA::Reader* reader_py = new TMVA::Reader("!Color:!Silent");

reader_py->AddVariable("var1",&elost00xlad00);
reader_py->AddVariable("var2",&elost00xlad01);
...
reader_py->AddVariable("var359",&elost19xlad16);
reader_py->AddVariable("var360",&elost19xlad17);

//Reading CNN models
  reader_py->BookMVA("PyKeras", "mod1.weights.xml");
  reader_py->BookMVA("PyKeras2", "mod2.weights.xml");     //Tried using different reader variables, but still crashes. Keeping same reader variable 

  //looping over event
  long int Nev = intree->GetEntries();
  for(long int iev=0; iev<Nev; iev++)
  {


	  pykeras = reader_py->EvaluateMVA("PyKeras");
        pykeras2 = reader_py->EvaluateMVA("PyKeras2");


outtree->Fill(); //Fill tree


  }

And here’s the crash message:

<WARNING> <WARNING>                : Failed to run python code: for i,p in enumerate(model.predict(vals, verbose=0)): output[i]=p
<WARNING> <WARNING>                : 
<WARNING> <WARNING>                : Python error message:
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/abayyari/.local/lib/python3.10/site-packages/keras/src/engine/training_v1.py", line 1059, in predict
    return func.predict(
  File "/home/abayyari/.local/lib/python3.10/site-packages/keras/src/engine/training_arrays_v1.py", line 801, in predict
    return predict_loop(
  File "/home/abayyari/.local/lib/python3.10/site-packages/keras/src/engine/training_arrays_v1.py", line 421, in model_iteration
    batch_outs = f(ins_batch)
  File "/home/abayyari/.local/lib/python3.10/site-packages/keras/src/backend.py", line 4609, in __call__
    fetched = self._callable_fn(*array_vals, run_metadata=self.run_metadata)
  File "/home/abayyari/.local/lib/python3.10/site-packages/tensorflow/python/client/session.py", line 1482, in __call__
    ret = tf_session.TF_SessionRunCallable(self._session._session,
tensorflow.python.framework.errors_impl.FailedPreconditionError: Could not find variable dense_1/bias. This could mean that the variable has been deleted. In TF1, it can also mean the variable is uninitialized. Debug info: container=localhost, status error message=Resource localhost/dense_1/bias/N10tensorflow3VarE does not exist.
	 [[{{node dense_1/BiasAdd/ReadVariableOp}}]]
<FATAL>                          : Failed to get predictions
***> abort program execution
terminate called after throwing an instance of 'std::runtime_error'
  what():  FATAL error

Any help would be greatly appreciated it!

2

Maybe @moneta can help