how to do train test split in keras imagedatagenerator

1train_datagen = ImageDataGenerator(rescale=1./255,
2    shear_range=0.2,
3    zoom_range=0.2,
4    horizontal_flip=True,
5    validation_split=0.2) # set validation split
6
7train_generator = train_datagen.flow_from_directory(
8    train_data_dir,
9    target_size=(img_height, img_width),
10    batch_size=batch_size,
11    class_mode='binary',
12    subset='training') # set as training data
13
14validation_generator = train_datagen.flow_from_directory(
15    train_data_dir, # same directory as training data
16    target_size=(img_height, img_width),
17    batch_size=batch_size,
18    class_mode='binary',
19    subset='validation') # set as validation data
20
21model.fit_generator(
22    train_generator,
23    steps_per_epoch = train_generator.samples // batch_size,
24    validation_data = validation_generator, 
25    validation_steps = validation_generator.samples // batch_size,
26    epochs = nb_epochs)
27