Skip to content

Scaling prototypes into ML models Questions

Practice questions for Scaling prototypes into ML models topic in Google Professional Machine Learning Engineer. 36 questions covering this domain.

36 questions8 easy18 medium10 hard
Q1
hard

A team configures a serverless training job to use TPU VMs and wants to scale a single worker pool to many replicas. Which constraint applies?

Q2
hard

A distributed multi-node GPU training job is spending too much time in all-reduce communication. Which Vertex AI optimization is intended to improve t...

Q3
medium

A team has a non-Python training application with private dependencies and a framework version that is not covered by Vertex AI prebuilt containers. W...

Q4
medium

An ML lead wants to compare a custom model and an AutoML model trained from the same managed dataset and also optionally export the resulting model in...

Q5
easy

A retailer wants to predict whether a customer will purchase a subscription. Which model type is the best fit?

Q6
easy

A finance team wants to predict next month's spend for each customer as a numeric amount. Which model type should they choose?

Q7
medium

A practitioner wants to optimize a system with expensive evaluations and many tunable parameters, and the system is not limited to machine learning. W...

Q8
medium

A data scientist creates a Vertex AI Vizier study and expects the service to execute each experiment automatically. What actually happens?

Q9
hard

A team wants to run a predictive model evaluation job in Vertex AI after training a custom tabular classifier. Which set of inputs is required for the...

Q10
medium

Which Vertex AI training feature is recommended when restartable training is required and a job can resume from intermediate state after preemption?

Q11
medium

A practitioner needs to scale training data for a deep learning model that no longer fits on a single VM. Which Vertex AI training feature should they...

Q12
medium

A team wants to use spot/preemptible compute to lower training cost while accepting that some jobs may be terminated. Which Vertex AI training option ...

Q13
medium

Which artifact does a Vertex AI hyperparameter tuning job report so that it can search over hyperparameters?

Q14
easy

Which open-source ML framework is supported by Vertex AI prebuilt training and serving containers?

Q15
hard

A team using AutoML Tabular notices the model overfits in the training pipeline. Which built-in capability of AutoML Tabular helps mitigate overfittin...

Q16
easy

Which Vertex AI service runs hyperparameter tuning trials by orchestrating training jobs with parameter suggestions from Vizier?

Q17
hard

A practitioner observes that custom training is bottlenecked by data input pipeline I/O while GPUs sit idle. Which optimization is most likely to impr...

Q18
medium

A team is choosing accelerators for a large transformer training job that benefits from very high interconnect bandwidth. Which option is generally re...

Q19
medium

A team is tuning a neural network and wants to stop unpromising trials early to save compute. Which Vertex AI hyperparameter tuning feature provides t...

Q20
medium

Which evaluation metric should a team use to assess overall performance of an object detection model across multiple IoU thresholds?

Sign in to see all 36 questions

Create a free account to browse all questions — completely free during our launch phase.