Keras Callbacks

Introduction

Callbacks provide hooks during training to monitor, stop, and save models at appropriate times.

Early Stopping

from tensorflow.keras.callbacks import EarlyStopping

early_stop = EarlyStopping(
    monitor='val_loss',
    patience=5,
    restore_best_weights=True,
    verbose=1
)

model.fit(x_train, y_train, epochs=50, callbacks=[early_stop])

Model Checkpoint

from tensorflow.keras.callbacks import ModelCheckpoint

checkpoint = ModelCheckpoint(
    'best_model.keras',
    monitor='val_accuracy',
    save_best_only=True,
    mode='max',
    verbose=1
)

model.fit(x_train, y_train, epochs=30, callbacks=[checkpoint])

TensorBoard

from tensorflow.keras.callbacks import TensorBoard

tensorboard = TensorBoard(
    log_dir='./logs',
    histogram_freq=1,
    write_graph=True,
    update_freq='epoch'
)

model.fit(x_train, y_train, epochs=10, callbacks=[tensorboard])

Custom Callback

class CustomCallback(keras.callbacks.Callback):
    def on_epoch_end(self, epoch, logs=None):
        if logs.get('accuracy') > 0.95:
            print(f"\nReached 95% accuracy at epoch {epoch}")
            self.model.stop_training = True

    def on_batch_end(self, batch, logs=None):
        print(f"\nBatch {batch}: loss = {logs.get('loss'):.4f}")

model.fit(x_train, y_train, epochs=10, callbacks=[CustomCallback()])

Learning Rate Scheduler

from tensorflow.keras.callbacks import LearningRateScheduler

def scheduler(epoch):
    if epoch < 10:
        return 0.001
    return 0.001 * 0.5 ** (epoch - 10)

lr_scheduler = LearningRateScheduler(scheduler, verbose=1)
model.fit(x_train, y_train, epochs=30, callbacks=[lr_scheduler])

Practice Problems

1. Use EarlyStopping to prevent overfitting
2. Save best model with ModelCheckpoint
3. Monitor training with TensorBoard
4. Create custom callback
5. Implement learning rate decay