How to use K.get_session in Tensorflow 2.0 or how to migrate it?

Question

def __init__(self, **kwargs):
    self.__dict__.update(self._defaults) # set up default values
    self.__dict__.update(kwargs) # and update with user overrides
    self.class_names = self._get_class()
    self.anchors = self._get_anchors()
    self.sess = K.get_session()

RuntimeError: get_session is not available when using TensorFlow 2.0.

Answer 1

Tensorflow 2.0 does not expose the backend.get_session directly any more but the code still there and expose for tf1.

https://github.com/tensorflow/tensorflow/blob/r2.0/tensorflow/python/keras/backend.py#L465

You can use it with tf1 compatible interface:

sess = tf.compat.v1.keras.backend.get_session() 

Or import tenforflow backend with internal path:

import tensorflow.python.keras.backend as K sess = K.get_session() 

Answer 2

In order to avoid using get_session after tensorflow 2.0 upgrade, Use tf.distribute.Strategy to get model. To load model, use tf.keras.models.load_model

import tensorflow as tf

another_strategy = tf.distribute.MirroredStrategy()
with another_strategy.scope():
    model = Service.load_deep_model()

def load_deep_model(self, model):
    loaded_model = tf.keras.models.load_model("model.h5")
    return loaded_model

Hope this helps. As this worked for me.

I have tried to explain same at this utility article as well. https://www.javacodemonk.com/runtimeerror-get_session-is-not-available-when-using-tensorflow-2-0-f7238546