Google Professional Machine Learning Engineer Question 253
Single answerGoogle Cloud PlatformYou are a Machine Learning Engineer tasked with training a custom NLP model on a large dataset. The model requires extensive preprocessing and custom TensorFlow training code. The team has access to Google Cloud and seeks a scalable and flexible training solution. Which of the following tools would be most appropriate for this task?
- A
Vertex AI custom training
- B
BigQuery ML
- C
AutoML Natural Language
- D
Kubeflow on Google Cloud
Show answer and explanation
Correct answer: A
Explanation
Vertex AI custom training is the most suitable option for this scenario because it allows you to run custom training code (e.g., TensorFlow), manage preprocessing, and scale effortlessly on Google Cloud. While Kubeflow offers flexibility, it introduces unnecessary complexity for this use case. BigQuery ML and AutoML Natural Language are not appropriate due to their lack of support for custom TensorFlow workflows.
- A. Correct.
Vertex AI custom training is ideal for custom training workflows where you can run your own training code (e.g., TensorFlow) and manage preprocessing steps, making it suitable for this scenario.
- B. Incorrect.
BigQuery ML is designed for training machine learning models using SQL and is not suitable for tasks requiring custom TensorFlow training code or extensive preprocessing.
- C. Incorrect.
AutoML Natural Language is a managed service for training NLP models without requiring custom code but lacks the flexibility needed for this custom TensorFlow training scenario.
- D. Incorrect.
Kubeflow on Google Cloud is a valid choice for orchestrating complex machine learning workflows, but it is more complex to manage compared to Vertex AI custom training for this specific use case.