test.py

# build and compile model

# length (in chars) of a label (this is a fixed value)
n = 7 
# total number (in chars) in label "alphabet": (0,1,2,3,4,5,6,7,8,9)
k = len("0123456789")

model = models.Sequential()
model.add(layers.Dense(x, activation='relu', input_shape=((N,))))
...
model.add(layers.Dense(n * k, activation=None))
model.add(layers.Reshape((n, k)))

model.add(layers.Dense(n, activation='softmax'))

model.compile(optimizer='rmsprop', 
             loss='categorical_crossentropy',
             metrics=['accuracy'])