Model is not predicting correct results keras tensorflow python

I am using tensorflow==1.15.5.

And I load my model like this

modelBlstm = tf.keras.models.load_model(modelPath,
                                                'elmo': hub.Module("", trainable=False),
                                                'tf': tf

My model needs elmo variable to be declared, so I passed the elmo variable in custom objects, but with this it does not predict correctly, now I declare the elmo as different variable

and just use

modelBlstm = tf.keras.models.load_model(modelPath)

it works well with the same user query.

But this works only if i am running the same code in single file.

But now if integrate the same code in my project it shows elmo variable is undefined.

Tried declaring elmo outside the class, creating elmo variable inside the class but it does not work, even with in the project, i removed the class and just cerated the functions like this…

dirName = os.path.dirname(__file__)

modelPath = os.path.join(dirName, "models/elmo-strat-bilstm-dense-2.h5")

modelBlstm = tf.keras.models.load_model(modelPath)

labelModelPath = os.path.join(dirName, "..\config\labels_mapping.pkl")

modelLabel = joblib.load(labelModelPath)

stopwords = ['i',  'am', 'a',

             'to', 'is', 'do', 'in', 'of']

elmo = hub.Module("", trainable=False)


with tf.Session() as session:


    modelBlstm = tf.keras.models.load_model(modelPath, )

def process(query: str) -> Intent:

    Logger.log("Processing query : " + query)

    query = __textPreprocess(query)

    print("query", query)

    intent = __predict(query)


    return intent

but even this does not work,

all other details i have mentioned in this SO question.

Any help?

Sorry, I’m replying but I don’t have an answer to you.

What I’d suggest to maybe help you is, if you want to use TF 2.x and with another text embedding, you can try this tutorial: Classify text with BERT  |  Text  |  TensorFlow