Callbacks

Code

import tensorflow as tf

callbacks = [
    tf.keras.callbacks.EarlyStopping(
        monitor="val_loss", patience=3, restore_best_weights=True),
    tf.keras.callbacks.ModelCheckpoint(
        "best.keras", monitor="val_loss", save_best_only=True),
    tf.keras.callbacks.ReduceLROnPlateau(
        monitor="val_loss", factor=0.5, patience=2, min_lr=1e-6),
    tf.keras.callbacks.TensorBoard(log_dir="./logs", histogram_freq=1),
    tf.keras.callbacks.CSVLogger("training.csv"),
]

model = tf.keras.Sequential([
    tf.keras.layers.Dense(32, activation="relu", input_shape=(10,)),
    tf.keras.layers.Dense(1, activation="sigmoid"),
])
model.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])

# Custom callback
class PrintEpoch(tf.keras.callbacks.Callback):
    def on_epoch_end(self, epoch, logs=None):
        print(f"epoch {epoch}: {logs}")

# model.fit(x_train, y_train, validation_split=0.2,
#           epochs=50, callbacks=callbacks + [PrintEpoch()])

Explanation

Callbacks hook into the training lifecycle to stop early, checkpoint the best weights, reduce the learning rate, or log to TensorBoard and CSV. EarlyStopping with restore_best_weights reverts to the best epoch instead of the last one. Subclassing Callback lets you run custom logic at any phase of training.

More TensorFlow Snippets