TensorFlow tf.data Pipelines 2

Question 1

Which tf.data method is used to read TFRecord files?

Accepted Answer

tf.data.TFRecordDataset()

Answer

TFRecordDataset reads TFRecord binary files, the recommended format for large-scale TensorFlow datasets.

Question 2

What does the num_parallel_calls parameter in dataset.map() do?

Accepted Answer

Parallelizes the map transformation across multiple CPU threads

Answer

num_parallel_calls allows multiple map function calls to run simultaneously, improving throughput on multi-core systems.

Question 3

Which tf.data method caches dataset elements to memory or disk?

Accepted Answer

dataset.cache()

Answer

dataset.cache() saves elements after the first epoch so subsequent epochs read from cache instead of re-processing.

Question 4

What does tf.io.parse_single_example() do?

Accepted Answer

Parses a single serialized tf.train.Example protobuf

Answer

tf.io.parse_single_example() deserializes a TFRecord's tf.train.Example into tensors using a feature description dict.

Question 5

Which method creates a dataset from a list of filenames?

Accepted Answer

tf.data.Dataset.list_files()

Answer

tf.data.Dataset.list_files() creates a dataset of file path strings matching a glob pattern.

TensorFlow Practice Test