TensorFlow Saving and Loading Models 2

Question 1

What is the default save format when calling model.save('my_model') without a file extension in TensorFlow 2.x?

Accepted Answer

SavedModel directory format

Answer

HDF5 format

Answer

Pickle format

Answer

Protocol Buffer frozen graph

Question 2

Which ModelCheckpoint argument ensures only the best-performing checkpoint is kept on disk?

Accepted Answer

save_best_only=True

Answer

best_only=True

Answer

monitor_best=True

Answer

keep_best=True

Question 3

What is the primary role of tf.train.CheckpointManager?

Accepted Answer

Limiting the number of checkpoints kept with max_to_keep

Answer

Optimizing checkpoint write speed

Answer

Distributing checkpoints across GPUs

Answer

Validating checkpoint file integrity

Question 4

Which method on tf.train.Checkpoint restores saved variable values?

Accepted Answer

.restore()

Answer

.load()

Answer

.import_variables()

Answer

.recover()

Question 5

What does the save_freq parameter in ModelCheckpoint control?

Accepted Answer

How often (in epochs or batches) a checkpoint is written

Answer

The learning rate at which weights are saved

Answer

The compression ratio of saved files

Answer

How many total checkpoints are allowed

Question 6

What must be true about a model before calling model.load_weights() successfully?

Accepted Answer

Its layer shapes must match those of the saved weights

Answer

It must use the identical optimizer

Answer

It must have been trained for the same number of epochs

Answer

It must use the same batch size

Question 7

Which low-level TensorFlow function explicitly saves a model in SavedModel format?

Accepted Answer

tf.saved_model.save()

Answer

tf.saved_model.export()

Answer

tf.saved_model.convert()

Answer

tf.saved_model.serialize()