Keras takes long time to go to next epoch

Question

While training the CNN, after completion of each epoch it takes more time to move to next epoch, while each epoch can be completed in 60s-80s, it take almost 5 mins to move to next epoch. I have provided my code, is there anything I am missing out ?

#importing the libraries

from keras.models import Sequential
from keras.layers import Conv2D
from keras.layers import MaxPooling2D
from keras.layers import Flatten
from keras.layers import Dense

#inintializing the ANN

classifier = Sequential()

# Convolutional layer

classifier.add(Conv2D(64,(3,3),input_shape =(128, 128, 3), activation = 'relu'))

#pooling layer
classifier.add(MaxPooling2D(pool_size = (2,2)))

#second convolutional layer
classifier.add(Conv2D(128,(3,3), activation = 'relu'))
classifier.add(MaxPooling2D(pool_size = (2,2)))

# flatten
classifier.add(Flatten())

#full connection
classifier.add(Dense(output_dim = 128, activation = 'relu'))

classifier.add(Dense(output_dim = 1, activation = 'sigmoid'))

#compiling the cnn
classifier.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'])

# we create two instances with the same arguments
from keras.preprocessing.image import ImageDataGenerator


train_datagen = ImageDataGenerator(rescale = 1./255,
                                   shear_range = 0.2,
                                   zoom_range = 0.2,
                                   horizontal_flip = True)

test_datagen = ImageDataGenerator(rescale = 1./255)

training_set = train_datagen.flow_from_directory('dataset/training_set',
                                                 target_size = (128, 128),
                                                 batch_size = 32,
                                                 class_mode = 'binary')

test_set = test_datagen.flow_from_directory('dataset/test_set',
                                            target_size = (128, 128),
                                            batch_size = 32,
                                            class_mode = 'binary')

classifier.fit_generator(training_set,
                         samples_per_epoch = 8000,
                         nb_epoch = 25,
                         validation_data = test_set,
                         nb_val_samples = 2000)

Answer 1

There is no need to set samples_per_epoch and nb_val_samples if you use ImageDataGenerator, as this is a Sequence and internally contains its length (if you use a recent Keras version, of course). The problem is that nb_val_steps is used for the parameter validation_steps, and I think you just set this value much higher than the right value.

If needed, you should set steps_per_epoch and validation_steps to the correct values, if you set validation_steps to a value larget than len(val_data) / batch_size, you are effectively telling keras to do validation with more data than necessary, slowing down the validation step.