Keras model load_weights for Neural Net

Question

I'm using the Keras library to create a neural network in python. I have loaded the training data (txt file), initiated the network and "fit" the weights of the neural network. I have then written code to generate the output text. Here is the code:

#!/usr/bin/env python

# load the network weights
filename = "weights-improvement-19-2.0810.hdf5"
model.load_weights(filename)
model.compile(loss='categorical_crossentropy', optimizer='adam')

My problem is: on execution the following error is produced:

 model.load_weights(filename)
 NameError: name 'model' is not defined

I have added the following but the error still persists:

from keras.models import Sequential
from keras.models import load_model

Any help would be appreciated.

Answer 1

you need to first create the network object called model, compile it and only after call the model.load_weights(fname)

working example:

from keras.models import Sequential
from keras.layers import Dense, Activation


def build_model():
    model = Sequential()

    model.add(Dense(output_dim=64, input_dim=100))
    model.add(Activation("relu"))
    model.add(Dense(output_dim=10))
    model.add(Activation("softmax"))

    # you can either compile or not the model
    model.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])
    return model


model1 = build_model()
model1.save_weights('my_weights.model')


model2 = build_model()
model2.load_weights('my_weights.model')

# do stuff with model2 (e.g. predict())

Save & Load an Entire Model

in Keras we can save & load the entire model like this (more info here):

from keras.models import load_model

model1 = build_model()
model1.save('my_model.hdf5')

model2 = load_model('my_model.hdf5')
# do stuff with model2 (e.g. predict()