Callbacks (EarlyStopping, ModelCheckpoint) in TensorFlow - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import EarlyStopping from TensorFlow Keras callbacks.

TensorFlow

from tensorflow.keras.callbacks import [1]

Drag options to blanks, or click blank then click option'

AModelCheckpoint

BTensorBoard

CReduceLROnPlateau

DEarlyStopping

Attempts:

3 left

💡 Hint

Common Mistakes

Importing ModelCheckpoint instead of EarlyStopping.

Trying to import from tensorflow.keras.models instead of callbacks.

EarlyStopping is imported from tensorflow.keras.callbacks to stop training early when a monitored metric stops improving.

2fill in blank

medium

Complete the code to create an EarlyStopping callback that monitors validation loss and stops if it doesn't improve for 3 epochs.

TensorFlow

early_stop = EarlyStopping(monitor='[1]', patience=3)

Drag options to blanks, or click blank then click option'

Aaccuracy

Bloss

Cval_loss

Dval_accuracy

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'loss' instead of 'val_loss' which monitors training loss.

Using 'accuracy' which is not the loss metric.

We monitor 'val_loss' to stop training when validation loss stops improving.

3fill in blank

hard

Fix the error in the code to save the best model weights only during training.

TensorFlow

checkpoint = ModelCheckpoint('best_model.h5', save_best_only=[1])

Drag options to blanks, or click blank then click option'

AFalse

BTrue

C'yes'

D1

Attempts:

3 left

💡 Hint

Common Mistakes

Using string 'yes' instead of boolean True.

Using integer 1 which may work but is not recommended.

The argument save_best_only expects a boolean True to save only the best model.

4fill in blank

hard

Fill both blanks to create a ModelCheckpoint that saves the model only when validation accuracy improves.

TensorFlow

checkpoint = ModelCheckpoint('model.h5', monitor='[1]', save_best_only=[2])

Drag options to blanks, or click blank then click option'

Aval_accuracy

Bval_loss

CTrue

DFalse

Attempts:

3 left

💡 Hint

Common Mistakes

Monitoring 'val_loss' instead of 'val_accuracy'.

Setting save_best_only to False which saves every epoch.

We monitor 'val_accuracy' and set save_best_only to True to save only the best model based on validation accuracy.

5fill in blank

hard

Fill all three blanks to create an EarlyStopping callback that monitors validation loss, waits 5 epochs before stopping, and restores the best weights.

TensorFlow

early_stop = EarlyStopping(monitor='[1]', patience=[2], restore_best_weights=[3])

Drag options to blanks, or click blank then click option'

Aval_loss

Bval_accuracy

C3

D5

ETrue

FFalse

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'val_accuracy' instead of 'val_loss'.

Setting patience to 3 instead of 5.

Not restoring best weights by setting restore_best_weights to False.

We monitor 'val_loss', set patience to 5 epochs, and restore_best_weights to True to keep the best model weights after stopping.