Which factor is most important when deciding between using CPUs and GPUs for a machine learning training job?

The type of operations in the model (e.g., matrix multiplications vs. data I/O)

The region where the training job will run

The number of CPU cores available

The storage type (SSD vs. HDD) used for training data

A team is training a very large model that does not fit in the memory of a single device. They are considering model parallelism. What is a key trade-off they should be aware of?

Model parallelism reduces memory usage but increases communication overhead

Model parallelism has no impact on memory or communication

Model parallelism reduces both memory usage and communication overhead

Model parallelism increases memory usage but reduces communication overhead

A team is training a model on a cluster of 8 GPUs using data parallelism. They notice that the training is not scaling well. They suspect the communication overhead is too high. Which technique can help reduce communication overhead in data parallelism?

Increase the batch size per GPU

Google CloudProfessional Machine Learning Engineer

Domain 3Objective 3

Choosing Appropriate Hardware for Training PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 3)

Part of the Scaling prototypes into ML models domain, which accounts for ~21% of the PROFESSIONAL-MACHINE-LEARNING-ENGINEER exam.

20questions here

4free pages

5concepts

~21%of the exam

Questions 11–15

11application · medium
A team is training a small natural language processing model on a dataset of 1 million sentences. They have a limited budget and need to train the model overnight. The model is a simple LSTM with 2 layers. Which compute option is most cost-effective for this workload?
Select an answer first
12foundation · easy
Which factor is most important when deciding between using CPUs and GPUs for a machine learning training job?
Select an answer first
13expert · hard
A team is training a very large model that does not fit in the memory of a single device. They are considering model parallelism. What is a key trade-off they should be aware of?
Select an answer first
14application · medium
A team is training a model on a cluster of 8 GPUs using data parallelism. They notice that the training is not scaling well. They suspect the communication overhead is too high. Which technique can help reduce communication overhead in data parallelism?
Select an answer first
15foundation · easy
Which GPU type is generally best suited for training a large language model that requires high memory capacity and fast inter-GPU communication?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to PROFESSIONAL-MACHINE-LEARNING-ENGINEER

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Google Cloud. “PROFESSIONAL-MACHINE-LEARNING-ENGINEER” is a trademark of its owner, used for identification only.

Choosing Appropriate Hardware for Training PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 3)

Questions 11–15

A team is training a small natural language processing model on a dataset of 1 million sentences. They have a limited budget and need to train the model overnight. The model is a simple LSTM with 2 layers. Which compute option is most cost-effective for this workload?

Which factor is most important when deciding between using CPUs and GPUs for a machine learning training job?

A team is training a very large model that does not fit in the memory of a single device. They are considering model parallelism. What is a key trade-off they should be aware of?

A team is training a model on a cluster of 8 GPUs using data parallelism. They notice that the training is not scaling well. They suspect the communication overhead is too high. Which technique can help reduce communication overhead in data parallelism?

Which GPU type is generally best suited for training a large language model that requires high memory capacity and fast inter-GPU communication?

Keep studying — Scaling prototypes into ML models