Issues When Applying Model.predict() Inside Data Generator With Fit_generator()

Basically I am implementing a model that employs perceptual loss to perform single image super-resolution. I constructed my full model such that the input will first pass through t

Solution 1:

The problem here is mutli-threading. When you are calling the 6 workers, block2_conv2/Relu:0 is created after graph is terminated.

The problem is with _make_predict_function(). You can check this file in your PC for the reasons(i got this from your error text) File "/home/lucien/anaconda3/envs/fyp/lib/python3.6/site-packages/keras/engine/training.py", line 1273, in predict_on_batch self._make_predict_function().

Some ways in which you can remove the errors are :

• Use theano backend.
• call model._make_predict_function() right after loading the trained model.
• Use global model :

Functions :

defload_model():
    global model
    model = yourmodel(weights=xx111122)
        # this is key : save the graph after loading the modelglobal graph
    graph = tf.get_default_graph()

While predicting:

with graph.as_default():
   preds = model.predict(image)
   #... etc