nerdexam
Google

PROFESSIONAL-MACHINE-LEARNING-ENGINEER · Question #209

You developed a Transformer model in TensorFlow to translate text. Your training data includes millions of documents in a Cloud Storage bucket. You plan to use distributed training to reduce…

The correct answer is A. Create a Vertex AI custom training job with GPU accelerators for the second worker pool. Use. To configure a distributed TensorFlow Transformer training job on Cloud Storage data with minimal code modification and cluster management effort, use a Vertex AI custom training job with GPU accelerators and tf.distribute.MultiWorkerMirroredStrategy.

Submitted by katya_ua· Apr 18, 2026ML model development

Question

You developed a Transformer model in TensorFlow to translate text. Your training data includes millions of documents in a Cloud Storage bucket. You plan to use distributed training to reduce training time. You need to configure the training job while minimizing the effort required to modify code and to manage the cluster's configuration. What should you do?

Options

  • ACreate a Vertex AI custom training job with GPU accelerators for the second worker pool. Use
  • BCreate a Vertex AI custom distributed training job with Reduction Server. Use N1 high-memory
  • CCreate a training job that uses Cloud TPU VMs. Use tf.distribute.TPUStrategy for distribution.
  • DCreate a Vertex AI custom training job with a single worker pool of A2 GPU machine type

How the community answered

(35 responses)
  • A
    74% (26)
  • B
    17% (6)
  • C
    3% (1)
  • D
    6% (2)

Why each option

To configure a distributed TensorFlow Transformer training job on Cloud Storage data with minimal code modification and cluster management effort, use a Vertex AI custom training job with GPU accelerators and `tf.distribute.MultiWorkerMirroredStrategy`.

ACreate a Vertex AI custom training job with GPU accelerators for the second worker pool. UseCorrect

Creating a Vertex AI custom training job allows you to leverage managed infrastructure for distributed training with minimal cluster management effort. Utilizing GPU accelerators for worker pools and implementing `tf.distribute.MultiWorkerMirroredStrategy` is the standard and recommended approach for synchronous distributed TensorFlow training, requiring minimal code modifications for existing models.

BCreate a Vertex AI custom distributed training job with Reduction Server. Use N1 high-memory

Using Reduction Server is typically for more advanced or custom distributed training setups and might not be the most minimal-effort approach for a standard TensorFlow Transformer model compared to `MultiWorkerMirroredStrategy`.

CCreate a training job that uses Cloud TPU VMs. Use tf.distribute.TPUStrategy for distribution.

While Cloud TPU VMs are powerful for Transformer models, `tf.distribute.TPUStrategy` often requires more specific code changes and optimization compared to `MultiWorkerMirroredStrategy` on GPUs, potentially increasing the effort for code modification.

DCreate a Vertex AI custom training job with a single worker pool of A2 GPU machine type

A 'single worker pool' configuration does not enable distributed training, which directly contradicts the requirement to 'use distributed training to reduce training time'.

Concept tested: Vertex AI custom training with TensorFlow distributed strategies

Source: https://cloud.google.com/vertex-ai/docs/training/distributed-training#tensorflow-multimachine-mirrored

Topics

#Distributed Training#Vertex AI Custom Training#TensorFlow tf.distribute#GPU Acceleration

Community Discussion

No community discussion yet for this question.

Sign up to join the discussion
Full PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice