Amazon MLS-C01 Exam - Topic 3 Question 111 Discussion

Actual exam question for Amazon's MLS-C01 exam

Question #: 111
Topic #: 3

An ecommerce company wants to train a large image classification model with 10.000 classes. The company runs multiple model training iterations and needs to minimize operational overhead and cost. The company also needs to avoid loss of work and model retraining.

Which solution will meet these requirements?

ACreate the training jobs as AWS Batch jobs that use Amazon EC2 Spot Instances in a managed compute environment.

BUse Amazon EC2 Spot Instances to run the training jobs. Use a Spot Instance interruption notice to save a snapshot of the model to Amazon S3 before an instance is terminated.

CUse AWS Lambda to run the training jobs. Save model weights to Amazon S3.

DUse managed spot training in Amazon SageMaker. Launch the training jobs with checkpointing enabled.

Show Suggested Answer

Suggested Answer: D

Amazon SageMaker managed spot training allows for cost-effective training by utilizing Spot Instances, which are lower-cost EC2 instances that can be interrupted when demand is high. By enabling checkpointing in SageMaker, the company can save intermediate model states to Amazon S3, allowing training to resume from the last checkpoint if interrupted. This solution minimizes operational overhead by automating the checkpointing process and resuming work after interruptions, reducing the need for retraining from scratch.

This setup provides a reliable and cost-efficient approach to training large models with minimal operational overhead and risk of data loss.

by Lavonne at Jan 24, 2025, 11:23 PM

Limited Time Offer

25%

Off

Get Premium MLS-C01 Questions as Interactive Web-Based Practice Test or PDF

Contribute your Thoughts:

Submit Cancel

Chery

3 months ago

A is good too, but I’d stick with D for less hassle.

upvoted 0 times

...

Shasta

3 months ago

Not sure about B, saving snapshots sounds risky.

upvoted 0 times

...

Ty

3 months ago

Wait, can AWS Lambda really handle large model training?

upvoted 0 times

...

Rodolfo

4 months ago

I agree, managed spot training is efficient!

upvoted 0 times

...

Omer

4 months ago

Option D seems like the best choice with checkpointing.

upvoted 0 times

...

Elvis

4 months ago

I’m leaning towards option B since it talks about saving a snapshot before termination, but I wonder if that’s the best way to handle interruptions.

upvoted 0 times

...

Karl

4 months ago

I feel like AWS Batch jobs could help with managing multiple training jobs, but I can't recall if they specifically address minimizing operational overhead.

upvoted 0 times

...

Pilar

4 months ago

I think option D sounds familiar because it mentions checkpointing, which we practiced in a similar question about model training.

upvoted 0 times

...

Emerson

5 months ago

I remember studying about using Spot Instances for cost efficiency, but I'm not sure if just using them alone is enough to avoid losing work.

upvoted 0 times

...

Shayne

5 months ago

I'm a bit confused by all the AWS service options. I'll need to review the details of each one to determine which one best fits the requirements. I don't want to rush into an answer and get it wrong.

upvoted 0 times

...

Diane

5 months ago

This seems straightforward to me. Option B, using EC2 Spot Instances with checkpointing to S3, is the way to go. It meets all the requirements and is a cost-effective solution.

upvoted 0 times

...

Christene

5 months ago

Okay, I think I've got a strategy for this. The key is to use a service that can handle the large number of classes and provide fault tolerance to avoid retraining. I'm leaning towards option D, managed spot training in Amazon SageMaker.

upvoted 0 times

...

Latrice

5 months ago

Hmm, I'm a bit unsure about this one. The question mentions a lot of different AWS services, so I'll need to make sure I understand how they work and which one would be the best fit.

upvoted 0 times

...

Jospeh

5 months ago

This looks like a tricky one. I'll need to carefully consider the requirements around minimizing operational overhead, avoiding loss of work, and the need to retrain the model.

upvoted 0 times

...

Josefa

1 year ago

Hold up, are we sure we can't just train this model on a single beefy EC2 instance? I mean, how hard can 10,000 classes be, right? *laughs nervously*

upvoted 0 times

Salena

12 months ago

B) Use Amazon EC2 Spot Instances to run the training jobs. Use a Spot Instance interruption notice to save a snapshot of the model to Amazon S3 before an instance is terminated.

upvoted 0 times

...

Galen

12 months ago

A) Create the training jobs as AWS Batch jobs that use Amazon EC2 Spot Instances in a managed compute environment.

upvoted 0 times

...

Mari

1 year ago

I'm putting my money on C. Using Lambda to run the training jobs and saving the model weights to S3? That's some next-level cloud-native goodness right there.

upvoted 0 times

Victor

11 months ago

D) Use managed spot training in Amazon SageMaker. Launch the training jobs with checkpointing enabled.

upvoted 0 times

...

Thaddeus

12 months ago

C) Use AWS Lambda to run the training jobs. Save model weights to Amazon S3.

upvoted 0 times

...

Salena

12 months ago

B) Use Amazon EC2 Spot Instances to run the training jobs. Use a Spot Instance interruption notice to save a snapshot of the model to Amazon S3 before an instance is terminated.

upvoted 0 times

...

Buddy

12 months ago

A) Create the training jobs as AWS Batch jobs that use Amazon EC2 Spot Instances in a managed compute environment.

upvoted 0 times

...

Solange

1 year ago

That's true, SageMaker could also be a good option for minimizing operational overhead.

upvoted 0 times

...

Latonia

1 year ago

Hmm, I'm torn between B and D. Saving the model to S3 before instances get terminated is a nice safety net, but the SageMaker option seems more seamless. Decisions, decisions...

upvoted 0 times

Lauran

12 months ago

Agreed. It's a tough choice, but either way, the company needs to prioritize minimizing operational overhead and cost.

upvoted 0 times

...

Mitsue

12 months ago

True, both options have their benefits. It really depends on what the company values more - cost efficiency or seamless operation.

upvoted 0 times

...

Buck

12 months ago

But B also has its advantages. Saving a snapshot of the model to S3 before termination can prevent loss of work.

upvoted 0 times

...

Laura

12 months ago

I think D is the way to go. Managed spot training in SageMaker with checkpointing sounds efficient.

upvoted 0 times

...

Zack

1 year ago

Option D definitely seems like the way to go. Managed spot training in SageMaker with checkpointing? That's the perfect combo for this use case. Gotta love that AWS magic.

upvoted 0 times

Justine

1 year ago

I agree, using managed spot training in Amazon SageMaker with checkpointing enabled would definitely help minimize operational overhead and cost while avoiding loss of work and model retraining.

upvoted 0 times

...

Keneth

1 year ago

Option D definitely seems like the way to go. Managed spot training in SageMaker with checkpointing? That's the perfect combo for this use case. Gotta love that AWS magic.

upvoted 0 times

...