Keras callbacks

Introduction to Deep Learning with Keras

Miguel Esteban

Data Scientist & Founder

What is a callback?

Callbacks in Keras

A callback you've been missing

# Training a model and saving its history
history = model.fit(X_train, y_train,
                    epochs=100,
                    metrics=['accuracy'])

print(history.history['loss'])

[0.6753975939750672, ..., 0.3155936544282096]

print(history.history['accuracy'])

[0.7030952412741525, ..., 0.8604761900220599]

A callback you've been missing

# Training a model and saving its history
history = model.fit(X_train, y_train,
                    epochs=100,
                    validation_data=(X_test, y_test),
                    metrics=['accuracy'])        

print(history.history['val_loss'])

[0.7753975939750672, ..., 0.4155936544282096]

print(history.history['val_accuracy'])

[0.6030952412741525, ..., 0.7604761900220599]

History plots

# Plot train vs test accuracy per epoch
plt.figure()

# Use the history metrics
plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])

# Make it pretty
plt.title('Model accuracy')
plt.ylabel('Accuracy')
plt.xlabel('Epoch')
plt.legend(['Train', 'Test'])
plt.show()

History plots

Early stopping

# Import early stopping from keras callbacks
from tensorflow.keras.callbacks import EarlyStopping


# Instantiate an early stopping callback
early_stopping = EarlyStopping(monitor='val_loss', patience=5)


# Train your model with the callback
model.fit(X_train, y_train, epochs=100,
          validation_data=(X_test, y_test),
          callbacks = [early_stopping])

Model checkpoint

# Import model checkpoint from keras callbacks
from keras.callbacks import ModelCheckpoint


# Instantiate a model checkpoint callback
model_save = ModelCheckpoint('best_model.hdf5',
                              save_best_only=True)


# Train your model with the callback
model.fit(X_train, y_train, epochs=100,
          validation_data=(X_test, y_test),
          callbacks = [model_save])

Let's practice!

Introduction to Deep Learning with Keras