Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Cloud Digital Leader practice bank has an additional 3,070 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
A retail company is training a large TensorFlow image recognition model. Its training jobs run for days on general-purpose CPU-based virtual machines, which slows down experimentation. The team wants specialized Google hardware that is purpose-built to speed up machine learning workloads.
Which Google Cloud option should the company choose to accelerate this training?
Terraform is an infrastructure-as-code tool. Google provides scripts that use it to deploy monitoring and logging dashboards that collect Cloud TPU metrics and logs on a single page in the Google Cloud console.
Dashboards make existing TPU data easier to view and interpret. They do not provide compute resources, and they do not accelerate model training. The company's challenge is slow training, which requires faster hardware.
The team had already moved to Cloud TPUs and wanted a single view of metrics and logs for debugging.
Tools like TensorBoard profiling can help optimize performance on TPUs. They support the accelerator but do not replace it as the source of the speed improvement.
Spot VMs are a consumption option offering lower-cost, preemptible capacity. They can be shut down at any time, which makes them a good fit for lower priority, interruption-tolerant jobs.
Spot pricing changes how much you pay and how reliably capacity is available. It does not change the underlying hardware architecture. The team's problem is slow training on general-purpose CPUs, and moving to cheaper CPUs would leave that unchanged, or even add interruptions.
The main goal was lowering the cost of an interruption-tolerant workload rather than speeding up training.
Spot is also available for TPUs. The cost option and the hardware choice are separate decisions, and the hardware choice is what drives speed here.
Cloud Monitoring is an observability service that automatically collects metrics, such as CPU utilization, network usage, and TPU idle duration, and works with Cloud Logging to track events. It helps teams see how their workloads and VMs are behaving.
The team needs hardware that makes training itself run faster. Monitoring only reports on performance. It does not add compute power or change how quickly a model trains.
The company already ran training on TPUs and wanted to detect unhealthy VMs, view idle times, or troubleshoot an underperforming job.
Monitoring and logging are complementary tools used after a workload is running on accelerators. They help verify that it works as expected, but they do not accelerate training.
A Tensor Processing Unit (TPU) is a custom-developed application-specific integrated circuit (ASIC) built by Google specifically to accelerate machine learning workloads. Cloud TPU is the service that gives you access to these chips from Google Cloud, including through Compute Engine, Google Kubernetes Engine, and Vertex AI.
Of the options, only Cloud TPUs combine purpose-built ML hardware with tight TensorFlow optimization, which directly addresses the problem of slow training.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated