CODEWITHSUNDEEP
pythontensorflowkeras
A cup of tea
A cup of tea

Reputation: 402

Keras: Load multiple models and predict in different threads

I'm using Keras with tensorflow core. I want to load 2 different models in constructor, and then make predictions (on request) in different threads. I tried to load these models within tensorflow graph contexts, but it didn't work. My code:

from keras.models import load_model
from keras import Sequential

def __init__(self):
    self.graph_A = tf.Graph()
    with self.graph_A.as_default():
        self.model_A: Sequential = load_model('model_A_filename')

    self.graph_B = tf.Graph()
    with self.graph_B.as_default():
        self.model_B: Sequential = load_model('model_B_filename')


def predict_with_model_A(X):
    with self.graph_A.as_default():
        return self.model_A.predict(X)


def predict_with_model_B(X):
    with self.graph_B.as_default():
        return self.model_B.predict(X)

When I run the program, model A is loaded successfully. However I receive an error on loading model B:

TypeError: Cannot interpret feed_dict key as Tensor: Tensor     
Tensor("Placeholder:0", shape=(7626, 210), dtype=float32) is not an element 
of this graph.

Will be happy to hear, how to do it right. Thanks!

Upvotes: 2

Views: 4050

Answers (1)

Tamil Selvan S
Tamil Selvan S

Reputation: 698

Try to reset the graph everytime after processing. For keras use K.clear_session(). Use seperate sessions for every graph. 

class Model:
   @staticmethod
   def loadmodel(path):
        return loadmodel(path)

def ___init__(self, path):
   self.model = self.loadmodel(path)
   self.graph = tf.get_default_graph()

def predict(self, X):
    with self.graph.as_default():
        return self.model.predict(X)

model1 = Model('model1.h5')
model1.predict(test_data)

model2 = Model('model2.h5')
model2.predict(test_data)

Upvotes: 2

Related Questions