Solves ValueError (using same graph)

[krshrimali]

Solves: ValueError problem because of not saving the graph just after loading the pre-trained weights of ResNet50.

Error:
(at execution of model.predict(image))

ValueError: Tensor Tensor("fc1000/Softmax:0", shape=(?, 1000), dtype=float32) is not an element of this graph.

System details:
Ubuntu 18.04
Keras Version: 2.1.6
Tensorflow Version: 1.8.0
Python Version: Python 3.6.5

Reference: tensorflow/tensorflow#14356 (comment)

[jrosebr1]

Your PR will fail if the user is using a non-TF backend (such as Theano, CNTK, mxnet, etc.). If you're making TF specific checks you'll want to check and see if the user is actually using TF.