Fine-tuning approach in TensorFlow - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pre-trained model for fine-tuning.

TensorFlow

base_model = tf.keras.applications.MobileNetV2(input_shape=(224, 224, 3), include_top=False, weights=[1])

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

ANone

B'random'

C'imagenet'

D'cifar10'

Attempts:

3 left

💡 Hint

Common Mistakes

Using None will not load pre-trained weights.

Using 'random' or 'cifar10' are invalid for this argument.

The pre-trained weights from ImageNet are loaded by setting weights='imagenet'. This helps the model start with learned features.

2fill in blank

medium

Complete the code to freeze the base model layers during fine-tuning.

TensorFlow

base_model.[1] = False

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

Arequires_grad

Btrainable

Cfrozen

Dfrozen_layers

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'frozen' or 'frozen_layers' are not valid attributes.

Using 'requires_grad' is PyTorch syntax.

Setting base_model.trainable = False freezes the layers so they are not updated during training.

3fill in blank

hard

Fix the error in the code to add a global average pooling layer after the base model.

TensorFlow

x = base_model.output
x = tf.keras.layers.[1]()(x)

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

AGlobalAveragePooling2D

BMaxPooling2D

CAveragePooling2D

DGlobalMaxPooling2D

Attempts:

3 left

💡 Hint

Common Mistakes

Using MaxPooling layers changes the pooling method.

AveragePooling2D is not global and requires pool size.

GlobalAveragePooling2D reduces each feature map to a single value by averaging, which is common after convolutional base models.

4fill in blank

hard

Fill both blanks to compile the fine-tuned model with an optimizer and loss function.

TensorFlow

model.compile(optimizer=tf.keras.optimizers.[1](learning_rate=0.0001), loss='[2]', metrics=['accuracy'])

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

AAdam

BSGD

Ccategorical_crossentropy

Dbinary_crossentropy

Attempts:

3 left

💡 Hint

Common Mistakes

Using SGD optimizer may require different learning rate tuning.

Using 'binary_crossentropy' is for two-class problems.

Adam optimizer is popular for fine-tuning with a small learning rate. For multi-class classification, 'categorical_crossentropy' is used as loss.

5fill in blank

hard

Fill all three blanks to create a dictionary of training callbacks for early stopping and saving the best model.

TensorFlow

callbacks = [
    tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=[1]),
    tf.keras.callbacks.ModelCheckpoint(filepath='best_model.h5', save_best_only=[2], monitor='[3]')
]

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

A3

BTrue

Cval_loss

DFalse

Attempts:

3 left

💡 Hint

Common Mistakes

Setting patience too high or zero.

Using False for save_best_only saves all models.

Monitoring training loss instead of validation loss.

EarlyStopping patience is set to 3 epochs. ModelCheckpoint saves only the best model based on validation loss.