Google Exam Professional Machine Learning Engineer Topic 3 Question 91 Discussion

Actual exam question for Google's Professional Machine Learning Engineer exam

Question #: 91
Topic #: 3

[All Professional Machine Learning Engineer Questions]

You work for an organization that operates a streaming music service. You have a custom production model that is serving a "next song" recommendation based on a user's recent listening history. Your model is deployed on a Vertex Al endpoint. You recently retrained the same model by using fresh dat

a. The model received positive test results offline. You now want to test the new model in production while minimizing complexity. What should you do?

ACreate a new Vertex Al endpoint for the new model and deploy the new model to that new endpoint Build a service to randomly send 5% of production traffic to the new endpoint Monitor end-user metrics such as listening time If end-user metrics improve between models over time gradually increase the percentage of production traffic sent to the new endpoint.

BCapture incoming prediction requests in BigQuery Create an experiment in Vertex Al Experiments Run batch predictions for both models using the captured data Use the user's selected song to compare the models performance side by side If the new models performance metrics are better than the previous model deploy the new model to production.

CDeploy the new model to the existing Vertex Al endpoint Use traffic splitting to send 5% of production traffic to the new model Monitor end-user metrics, such as listening time If end-user metrics improve between models over time, gradually increase the percentage of production traffic sent to the new model.

DConfigure a model monitoring job for the existing Vertex Al endpoint. Configure the monitoring job to detect prediction drift, and set a threshold for alerts Update the model on the endpoint from the previous model to the new model If you receive an alert of prediction drift, revert to the previous model.

Show Suggested Answer

Suggested Answer: C

Traffic splitting is a feature of Vertex AI that allows you to distribute the prediction requests among multiple models or model versions within the same endpoint. You can specify the percentage of traffic that each model or model version receives, and change it at any time. Traffic splitting can help you test the new model in production without creating a new endpoint or a separate service. You can deploy the new model to the existing Vertex AI endpoint, and use traffic splitting to send 5% of production traffic to the new model. You can monitor the end-user metrics, such as listening time, to compare the performance of the new model and the previous model. If the end-user metrics improve between models over time, you can gradually increase the percentage of production traffic sent to the new model. This solution can help you test the new model in production while minimizing complexity and cost.Reference:

Traffic splitting | Vertex AI

Deploying models to endpoints | Vertex AI

by Selma at Sep 16, 2024, 08:52 PM

Limited Time Offer

25%

Off

Get Premium Professional Machine Learning Engineer Questions as Interactive Web-Based Practice Test or PDF

Contribute your Thoughts:

Submit Cancel

Nakita

2 months ago

I prefer option B. Capturing incoming prediction requests in BigQuery and running batch predictions for both models seems like a thorough approach to compare performance.

upvoted 0 times

...

Thomasena

2 months ago

I'm partial to B - I love a good data-driven experiment! Although, I hope they're not using the 'song selected' metric as the only KPI. Gotta look at the whole user experience.

upvoted 0 times

...

Tasia

2 months ago

Option C is the way to go - it's the 'Goldilocks' solution, not too disruptive, not too risky. Just right!

upvoted 0 times

...

Daren

2 months ago

A reminds me of A/B testing, which is a classic approach. I wonder if the random 5% split might lead to some users getting a suboptimal experience though.

upvoted 0 times

Tashia

29 days ago

A) I agree, A/B testing is a common method. But you're right, there could be some users who might not have the best experience with the 5% split.

upvoted 0 times

...

Isadora

1 months ago

C) Deploy the new model to the existing Vertex Al endpoint Use traffic splitting to send 5% of production traffic to the new model Monitor end-user metrics, such as listening time If end-user metrics improve between models over time, gradually increase the percentage of production traffic sent to the new model.

upvoted 0 times

...

Josue

1 months ago

B) Capture incoming prediction requests in BigQuery Create an experiment in Vertex Al Experiments Run batch predictions for both models using the captured data Use the user's selected song to compare the models performance side by side If the new models performance metrics are better than the previous model deploy the new model to production.

upvoted 0 times

...

Melynda

1 months ago

A) Create a new Vertex Al endpoint for the new model and deploy the new model to that new endpoint Build a service to randomly send 5% of production traffic to the new endpoint Monitor end-user metrics such as listening time If end-user metrics improve between models over time gradually increase the percentage of production traffic sent to the new endpoint.

upvoted 0 times

...

Vincenza

2 months ago

D is a clever idea, using monitoring to automatically update the model. But I'm not sure I'd trust the drift detection to work perfectly on the first try.

upvoted 0 times

Vince

1 months ago

D) Configure a model monitoring job for the existing Vertex AI endpoint. Configure the monitoring job to detect prediction drift, and set a threshold for alerts. Update the model on the endpoint from the previous model to the new model. If you receive an alert of prediction drift, revert to the previous model.

upvoted 0 times

...

Candra

1 months ago

C) Deploy the new model to the existing Vertex AI endpoint. Use traffic splitting to send 5% of production traffic to the new model. Monitor end-user metrics, such as listening time. If end-user metrics improve between models over time, gradually increase the percentage of production traffic sent to the new model.

upvoted 0 times

...

Arlean

2 months ago

A) Create a new Vertex AI endpoint for the new model and deploy the new model to that new endpoint. Build a service to randomly send 5% of production traffic to the new endpoint. Monitor end-user metrics such as listening time. If end-user metrics improve between models over time, gradually increase the percentage of production traffic sent to the new endpoint.

upvoted 0 times

...

Miesha

2 months ago

I agree with Lorrie. Option A seems like the most efficient way to test the new model while minimizing complexity.

upvoted 0 times

...

Lorrie

3 months ago

I think option A is the best choice. It allows for gradual testing of the new model in production.

upvoted 0 times

...

Ling

3 months ago

B looks interesting, but capturing all the prediction requests in BigQuery could get costly. I'd prefer a more targeted approach like C.

upvoted 0 times

...

Sheldon

3 months ago

Option C seems the most straightforward and minimizes disruption to the production environment. I like how it gradually ramps up the new model's traffic to monitor performance.

upvoted 0 times

Eun

2 months ago

Definitely. It's a smart approach to minimize disruption while testing the new model.

upvoted 0 times

...

Kirk

2 months ago

I think monitoring end-user metrics is key to ensuring the new model is performing well.

upvoted 0 times

...

Lavera

2 months ago

I agree. It's important to monitor performance before fully deploying the new model.

upvoted 0 times

...

Oretha

2 months ago

I agree. It's important to monitor the end-user metrics and gradually increase the traffic to the new model to ensure it performs well.

upvoted 0 times

...

Latricia

3 months ago

Option C seems like a good choice. It allows for gradual testing of the new model.

upvoted 0 times

...

Deangelo

3 months ago

Option C seems like the best approach. It allows for gradual testing of the new model without causing too much disruption.

upvoted 0 times

...