Tensorflow reuse neural network

Question

I'm new in tensorflow and I've been training a simple neural network, but once is trained, I don't know how to reuse the NN to get the outputs of an input.

def train_neural_network(x,y,aDataTrain,aTargetTrain,aDataTest,aTargetTest):
    batch_size = 500
    prediction = neural_network_model(x,len(aDataTrain[0]))
    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,labels=y))
    optimizer = tf.train.AdamOptimizer().minimize(cost)
    hm_epochs = 1

    saver = tf.train.Saver()
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())

        for epoch in range(hm_epochs):
            epoch_loss = 0
            i = 0
            while i < len(aDataTrain):
                start = i
                end = i + batch_size
                batch_x = np.array(aDataTrain[start:end])
                batch_y = np.array(aTargetTrain[start:end])

                _,c = sess.run([optimizer,cost],feed_dict={x:batch_x,y:batch_y})
                epoch_loss += c
                i += batch_size
            print ("Epoch", epoch, "completed out of", hm_epochs, "loss", epoch_loss)

        correct =tf.equal(tf.argmax(prediction,1), tf.argmax(y,1))

        accurracy = tf.reduce_mean(tf.cast(correct,'float'))
        finalAcc = accurracy.eval({x:aDataTest,y:aTargetTest})
        saver.save(sess, 'model/model.ckpt')

    print("Accuracy:",finalAcc)

So, once I've saved the model and try to restore it, I don't know how to continue to get the output of the NN from the "input_data".

def execute_neural_network(x,y,aDataTrain,aTargetTrain,aDataTest,aTargetTest):
    batch_size = 1
    y_pred = []

    prediction = neural_network_model(x,len(aDataTrain[0]))
    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,labels=y))
    optimizer = tf.train.AdamOptimizer().minimize(cost)

    input_data = [5.0, 3.0, 1.0, 5.0, 6.0, 5.0, 2.0, 4.0, 7.0, 6.0, 3.0, 3.0, 3.0, 3.0, 3.0, 4.0, 2.0, 3.0, 3.0, 3.0, 3.0, 3.0, 2.0, 3.0, 2.0, 3.0, 2.0, 3.0, 3.0, 4.0, 3.0, 3.0, 2.0, 4.0, 3.0, 3.0, 2.0, 4.0, 3.0, 3.0, 3.0, 3.0, 3.0, 3.0, 61.0, 21.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 75.0, 3.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 6.0, 35.0, 11.0, 10.0, 33.0, 24.0, 6.0, 2.0, 2.0, 3.0, 4.0, 3.0, 3.0, 8.0, 6.0, 5.0, 6.0, 5.0, 8.0, 9.0, 13.0, 7.0, 25.0, 11.0, 2.0, 2.0, 2.0, 2.0, 2.0]

    saver = tf.train.Saver()
    with tf.Session() as sess:
        saver.restore(sess, 'model/model.ckpt')
        #Get neural network output from input_data

zimmerrol · Answer 1 · 2017-11-09T07:44:59.590

Assuming you create your graph/network's model somehow like this:

with tf.Session() as sess:
    #do other stuff
    predictionOp = tf.argmax(py_x, 1)

saver.save(sess, 'model')

where predictionOp is the variable which is the output of your network.

You can add something like this afterwards: tf.add_to_collection("predictionOp", predictionOp) to give the predictionOp a name to be easier to find. Then, you can reload your model and get the predictions with:

with tf.Session() as sess:
    new_saver = tf.train.import_meta_graph('model.meta')
    new_saver.restore(sess, 'model')
    predictionOp = tf.get_collection("predictionOp")[0]

    #get the prediction
    prediction = sess.run(predictionOp, feed_dict={"x:0": input_data})

For more information, please take a look at the tensorflow documentation and here for some more information about the basics. Also, there are some other threads, which deal with similar problems, like this and this one.

Tensorflow reuse neural network

1 Answers1