Google Professional Machine Learning Engineer Question 408
Single answerGoogle Cloud PlatformYou are designing a machine learning pipeline using TensorFlow Extended (TFX) components on Google Cloud. Your pipeline includes steps for data ingestion, data validation, model training, and model deployment. The input data resides in Google Cloud Storage (GCS), and you want to ensure that the data processing scales efficiently with growing data volumes. Which TFX component or approach is best suited for the data ingestion step in this scenario?
- A
TFX ExampleGen using Apache Beam with Dataflow as the runner
- B
TFX Evaluator to process the data and validate it
- C
TFX Transform to apply data transformations directly to GCS files
- D
TFX Pusher to load data from GCS into the pipeline
Show answer and explanation
Correct answer: A
Explanation
In a TFX pipeline, ExampleGen is the component responsible for ingesting data into the pipeline. By configuring ExampleGen to use Apache Beam with Dataflow as the runner, you can efficiently process large datasets stored in GCS in a distributed and scalable manner. This ensures that the data ingestion step can handle increasing data volumes without becoming a bottleneck.
- A. Correct.
TFX ExampleGen is the correct choice for data ingestion. Using Apache Beam with Dataflow as the runner ensures scalable and distributed data processing for large datasets stored in GCS.
- B. Incorrect.
TFX Evaluator is used for model evaluation, not for data ingestion, so this is incorrect.
- C. Incorrect.
TFX Transform is used to apply data transformations, but it does not handle data ingestion directly from GCS.
- D. Incorrect.
TFX Pusher is responsible for deploying the trained model to a serving infrastructure. It is not used for data ingestion.