nerdexam
Google

PROFESSIONAL-MACHINE-LEARNING-ENGINEER · Question #220

You are training a custom language model for your company using a large dataset. You plan to use the Reduction Server strategy on Vertex AI. You need to configure the worker pools of the distributed…

The correct answer is B. Configure the machines of the first two worker pools to have GPUs and to use a container image. To configure worker pools for a Vertex AI distributed training job using the Reduction Server strategy for a custom language model, the machines in the computational worker pools should be equipped with GPUs and use a pre-configured container image with all necessary…

Submitted by asante_acc· Apr 18, 2026ML model development

Question

You are training a custom language model for your company using a large dataset. You plan to use the Reduction Server strategy on Vertex AI. You need to configure the worker pools of the distributed training job. What should you do?

Options

  • AConfigure the machines of the first two worker pools to have GPUs, and to use a container image
  • BConfigure the machines of the first two worker pools to have GPUs and to use a container image
  • CConfigure the machines of the first two worker pools to have TPUs and to use a container image
  • DConfigure the machines of the first two pools to have TPUs, and to use a container image where

How the community answered

(39 responses)
  • A
    3% (1)
  • B
    79% (31)
  • C
    10% (4)
  • D
    8% (3)

Why each option

To configure worker pools for a Vertex AI distributed training job using the Reduction Server strategy for a custom language model, the machines in the computational worker pools should be equipped with GPUs and use a pre-configured container image with all necessary distributed training dependencies.

AConfigure the machines of the first two worker pools to have GPUs, and to use a container image

Option A is textually identical to Option B, suggesting a potential duplication in the question's choices, but the technical reasoning for using GPUs and a container image is correct for distributed training.

BConfigure the machines of the first two worker pools to have GPUs and to use a container imageCorrect

For distributed training of a custom language model using the Reduction Server strategy, configuring the computational worker pools (such as the first two, typically chief and a worker) with GPUs is standard practice for accelerating deep learning workloads. Additionally, using a custom container image with all distributed training dependencies pre-installed is essential for reliable and efficient execution on Vertex AI.

CConfigure the machines of the first two worker pools to have TPUs and to use a container image

While TPUs are powerful accelerators, using them requires specific code adaptations for the model. Without explicit mention of such adaptations or a requirement for TPU-specific optimizations, GPUs are the more general and commonly used accelerators for custom language model training, making TPUs a less likely default for this scenario.

DConfigure the machines of the first two pools to have TPUs, and to use a container image where

Similar to option C, using TPUs necessitates specialized code changes for TensorFlow models to leverage them effectively. The 'Reduction Server strategy' itself doesn't inherently mandate TPUs, and GPUs are a versatile choice for distributed deep learning.

Concept tested: Distributed training worker pool configuration (GPUs/containers)

Source: https://cloud.google.com/vertex-ai/docs/training/create-custom-job#distributed-training

Topics

#Distributed Training#Vertex AI#GPUs#Custom Containers

Community Discussion

No community discussion yet for this question.

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