Google Professional Machine Learning Engineer Exam - Topic 1 Question 52 Discussion

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

Question #: 52
Topic #: 1

[All Professional Machine Learning Engineer Questions]

You are developing an ML model intended to classify whether X-Ray images indicate bone fracture risk. You have trained on Api Resnet architecture on Vertex AI using a TPU as an accelerator, however you are unsatisfied with the trainning time and use memory usage. You want to quickly iterate your training code but make minimal changes to the code. You also want to minimize impact on the models accuracy. What should you do?

AConfigure your model to use bfloat16 instead float32

BReduce the global batch size from 1024 to 256

CReduce the number of layers in the model architecture

DReduce the dimensions of the images used un the model

Show Suggested Answer

Suggested Answer: B

by Casie at Sep 21, 2022, 08:08 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

Jesus

4 months ago

A and B are solid choices for quick fixes!

upvoted 0 times

...

Novella

4 months ago

Wait, reducing image dimensions? Isn't that a big deal?

upvoted 0 times

...

Fausto

4 months ago

C seems risky, might lose important features.

upvoted 0 times

...

Antonio

4 months ago

I think B might hurt accuracy too much.

upvoted 0 times

...

Alton

4 months ago

A is a good option for faster training!

upvoted 0 times

...

Francesco

5 months ago

I think reducing image dimensions might help with memory, but I worry it could impact the model's ability to detect fractures. Option D seems risky.

upvoted 0 times

...

Val

5 months ago

I feel like reducing the number of layers could hurt accuracy too much, so I’m hesitant about option C.

upvoted 0 times

...

Malika

5 months ago

I'm not sure, but I think reducing the batch size could speed up training time. Option B sounds familiar from practice questions.

upvoted 0 times

...

Katina

5 months ago

I remember reading that using bfloat16 can help with memory usage without significantly affecting accuracy, so option A might be a good choice.

upvoted 0 times

...

Marylou

5 months ago

I'm a bit confused by the `Sequence` class. I'll need to make sure I understand how it's generating the sequence of numbers.

upvoted 0 times

...

Marlon

5 months ago

Ah, I see what's going on here. The witness components are flagged as absent, so the administrator needs to re-integrate the Witness Host into the cluster. Option D looks like the right choice to me.

upvoted 0 times

...

Tonja

5 months ago

Okay, let me break this down step-by-step. I know syslog messages are sent from the switches, but I'm not sure about the exact path. I'll need to review the syslog configuration in ACI to determine the correct answer.

upvoted 0 times

...

Peggie

5 months ago

Ah, this is a tricky one. I'm not super familiar with the details of the top-down SOA approach. I'll have to use the process of elimination to narrow down the options and make my best guess.

upvoted 0 times

...