nerdexam
Google

PROFESSIONAL-DATA-ENGINEER · Question #210

You are working on a niche product in the image recognition domain. Your team has developed a model that is dominated by custom C++ TensorFlow ops your team has implemented. These ops are used…

The correct answer is D. Stay on CPUs, and increase the size of the cluster you're training your model on. Explanation/Reference: CPUs Models that are dominated by custom TensorFlow operations written in C++ Cloud TPUs are not suited to the following workloads: Neural network workloads that contain custom TensorFlow operations written in C++. Specifically, custom operations in the…

Submitted by packet_pusher· Mar 30, 2026Operationalizing machine learning models

Question

You are working on a niche product in the image recognition domain. Your team has developed a model that is dominated by custom C++ TensorFlow ops your team has implemented. These ops are used inside your main training loop and are performing bulky matrix multiplications. It currently takes up to several days to train a model. You want to decrease this time significantly and keep the cost low by using an accelerator on Google Cloud. What should you do?

Options

  • AUse Cloud TPUs without any additional adjustment to your code.
  • BUse Cloud TPUs after implementing GPU kernel support for your customs ops.
  • CUse Cloud GPUs after implementing GPU kernel support for your customs ops.
  • DStay on CPUs, and increase the size of the cluster you're training your model on.

How the community answered

(30 responses)
  • A
    13% (4)
  • B
    23% (7)
  • C
    7% (2)
  • D
    57% (17)

Explanation

Explanation/Reference: CPUs Models that are dominated by custom TensorFlow operations written in C++ Cloud TPUs are not suited to the following workloads: Neural network workloads that contain custom TensorFlow operations written in C++. Specifically, custom operations in the body of the main training loop are not suitable for TPUs. https://cloud.google.com/tpu/docs/tpus

Topics

#Cloud TPUs#custom TensorFlow ops#GPU kernels#hardware accelerators

Community Discussion

No community discussion yet for this question.

Sign up to join the discussion
Full PROFESSIONAL-DATA-ENGINEER Practice