Keras model working fine locally but won't wor

问题:

I'm working on this heart disease detection problem using different classifiers. What I'm doing is i'm saving my model in h5 file & creating an object of it & returning the response in json format.

but the same model won't work on flask api that ran perfectly on my terminal.

here is my Neural network:

def ANN():
    global x_train,x_test,y_train,y_test
    model = Sequential()

    #implicit input layer combined with hidden layer
    model.add(Dense(units = 13, kernel_initializer = 'uniform', activation = 'relu', input_dim = 13))

    #hidden layer 2
    model.add(Dense(units = 13, kernel_initializer = 'uniform', activation = 'relu', input_dim = 13))

    #output layer
    model.add(Dense(units = 1, kernel_initializer = 'uniform', activation = 'sigmoid'))

    model.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'])

    #fitting with optimal hyperparameters
    model.fit(x_train, y_train, batch_size = 25, nb_epoch = 287)

    return {'model':model,
            'accuracy':accuracy_score(model.predict(x_test) > 0.5,y_test)*100}

after saving the model in .h5 file, In my flask api,

ann = load_model('ann8524.h5')
print(ann.predict(x_test)) #test set, for just checking.

here's the error message:

* Running on http://127.0.0.1:5000/ (Press CTRL+C to quit) [2018-12-20 23:37:43,548] 
ERROR in app: Exception on /heart/predict

[GET] Traceback (most recent call last):   
File "C:\python_installation\lib\site-packages\flask\app.py", line 1813, in full_dispatch_request
    rv = self.dispatch_request()   
File "C:\python_installation\lib\site-packages\flask\app.py", line 1799, in dispatch_request
    return self.view_functions[rule.endpoint](**req.view_args)   
File "C:\python_installation\lib\site-packages\flask_restful\__init__.py", line 458, in wrapper
    resp = resource(*args, **kwargs)   
File "C:\python_installation\lib\site-packages\flask\views.py", line 88, in view
    return self.dispatch_request(*args, **kwargs)   
File "C:\python_installation\lib\site-packages\flask_restful\__init__.py", line 573, in dispatch_request
    resp = meth(*args, **kwargs)   
File "app.py", line 41, in get
    print(ann.predict(x_test))   
File "C:\python_installation\lib\site-packages\keras\engine\training.py", line 1164, in predict
    self._make_predict_function()   
File "C:\python_installation\lib\site-packages\keras\engine\training.py", line 554, in _make_predict_function
    **kwargs)   
File "C:\python_installation\lib\site-packages\keras\backend\tensorflow_backend.py", line 2744, in function
    return Function(inputs, outputs, updates=updates, **kwargs)   
File "C:\python_installation\lib\site-packages\keras\backend\tensorflow_backend.py", line 2546, in __init__
    with tf.control_dependencies(self.outputs):   
File "C:\python_installation\lib\site-packages\tensorflow\python\framework\ops.py", line 5004, in control_dependencies
    return get_default_graph().control_dependencies(control_inputs)   
File "C:\python_installation\lib\site-packages\tensorflow\python\framework\ops.py", line 4543, in control_dependencies
    c = self.as_graph_element(c)   
File "C:\python_installation\lib\site-packages\tensorflow\python\framework\ops.py", line 3490, in as_graph_element
    return self._as_graph_element_locked(obj, allow_tensor, allow_operation)   
File "C:\python_installation\lib\site-packages\tensorflow\python\framework\ops.py", line 3569, in _as_graph_element_locked
    raise ValueError("Tensor %s is not an element of this graph." % obj) 

ValueError: Tensor Tensor("dense_3/Sigmoid:0", shape=(?, 1), dtype=float32) is not an element of this graph.

127.0.0.1 - - [20/Dec/2018 23:37:43] "[1m[35mGET /heart/predict HTTP/1.1[0m" 500 -

But it works perfectly fine in Spyder. (exact same code)

回答1:

You need to get default graph from Tensorflow, following these steps should solve this issue:

import tensorflow as tf
ann = load_model('ann8524.h5')
graph = tf.get_default_graph()

def your_handler():
    global graph
    with graph.as_default():
        print(ann.predict(x_test))