Load ML model as Pyobject? and pass it to another Interpret tp avoid loading again

[JepDev]

Good day folks,

I am trying to do a one time, import of tensorflow and  load of a ML model from python in Java. Then pass this model 

for predictions. For instance:

-------------

try (Interpreter interp = new SharedInterpreter()){

interp.exec("import tensorflow");

interp.exec("MLmodel= tf.load_model('someModel')");

// Not sure how to create an object of the model, maybe pyObject

} catch (JepException e) {

// TODO Auto-generated catch block

e.printStackTrace();

}

----------

// do some operations of data in java then apply the model to the java data

-------------

try (Interpreter interp = new SharedInterpreter()){

// I want to avoid loading the model and importing tensorflow again

interp.exec("x = somemodel(data)");

} catch (JepException e) {

// TODO Auto-generated catch block

e.printStackTrace();

}

-------

I am aware of pyObjects and pyCallable which seems like the way to go, but as I am a noobie at java, I would need a brief example to see how to implement it. 

Points to note:

I tried using:

--------

  JepConfig cfg = new JepConfig()

  .addSharedModules("numpy,tensorflow");

  try {

    SharedInterpreter.setConfig(cfg);

    

  } catch (JepException e) {

       e.printStackTrace();

  }

---------

to set the preloaded modules but sharedInterpreter doesn't have this functionality it seems. 

Can anyone help.

Thanks in advance,