Google Professional Machine Learning Engineer Question 251
Select 3Google Cloud PlatformYou are working on a machine learning project and need to train a custom model using TensorFlow. The model requires distributed training across multiple GPUs. You are deciding between using Vertex AI custom training and Kubeflow on Google Cloud. Which of the following factors should influence your decision?
- A
The need for built-in hyperparameter tuning support during training.
- B
The need for full control over the orchestration of distributed training jobs.
- C
The ability to use pre-built containers for faster setup.
- D
The requirement to use open-source frameworks and maintain portability across cloud providers.
- E
The need for seamless integration with Google Cloud’s managed services such as AI Platform.
Show answer and explanation
Correct answers: A, B, D
Explanation
When deciding between Vertex AI custom training and Kubeflow on Google Cloud, you should consider the specific needs of your ML project. Vertex AI is better suited for scenarios requiring built-in hyperparameter tuning and managed services. On the other hand, Kubeflow is ideal for projects needing full control over orchestration, open-source portability, and support for complex distributed training workflows. Choosing the right tool depends on the balance between control, ease of use, and portability.
- A. Correct.
Vertex AI custom training provides built-in hyperparameter tuning, which can simplify the training process and improve model performance.
- B. Correct.
Kubeflow offers full control over the orchestration of distributed training jobs, making it suitable for complex workflows requiring fine-grained configurations.
- C. Incorrect.
While both Vertex AI and Kubeflow support pre-built containers, this factor alone does not differentiate between the two for this scenario.
- D. Correct.
Kubeflow is an open-source platform that allows you to maintain portability across cloud providers, making it a better choice if this is a requirement.
- E. Incorrect.
Although Vertex AI integrates seamlessly with Google Cloud’s managed services, this factor does not directly address the requirements of distributed training or TensorFlow-specific needs.