Exam AI-100 topic 4 question 25 discussion

Actual exam question from Microsoft's AI-100

Question #: 25
Topic #: 4

DRAG DROP -
You create an image classification model in Azure Machine Learning Studio.
You need to deploy the model as a containerized web service.
Which four actions should you perform in sequence? To answer, move the appropriate actions from the list of actions to the answer area and arrange them in the correct order.
Select and Place:

Show Suggested Answer

Suggested Answer:

The following diagram illustrates the complete deployment workflow:

The deployment workflow includes the following steps:
1. Register the model in a registry hosted in your Azure Machine Learning Service workspace
2. Register an image that pairs a model with a scoring script and dependencies in a portable container
3. Deploy the image as a web service in the cloud or to edge devices
4. Monitor and collect data
5. Update a deployment to use a new image.
References:
https://docs.microsoft.com/bs-latn-ba/azure/machine-learning/service/concept-model-management-and-deployment#step-3-deploy-image

by jadepe at Sept. 17, 2020, 5:03 p.m.

Comments

Submit Cancel

jadepe

Highly Voted 4 years, 8 months ago

1 Create image 2 Register image 3 Start container 4 Ge endpoint The model is supposed to be already created and an image is just a template yo need to instantiate by running the container and then get the endpoint

upvoted 24 times

...

KenCraw

Most Recent 4 years, 3 months ago

Any chance it could be Train > Register > Start > Endpoint (as per the image)? Is "create" required, or could that be included in the "register" process?

upvoted 1 times

Cornholioz

4 years, 3 months ago

Well, I think create and register are two steps. But I see where you are coming from. You are looking to fit in the required actions within the 4 steps they've asked for. It doesn't say if the model was already trained. So, not sure if it is a required step. (But the image shows a "Trained" Model) Then again, Starting the Container is also required after the image is created & registered. Getting the endpoint is required because the question asks about deploying it as a webservice which means it is probably good to state this as a step in process. But it asks for just 4 steps. Poorly framed question!

upvoted 1 times

...

UpsetUser

4 years, 4 months ago

Given answer is correct as per V__t__Se___ Practice tests

upvoted 1 times

Cornholioz

4 years, 3 months ago

The practice test probably copied the answer from here or vice versa. If there is a correct answer as per documentation or an actual hands-on-keyboard, that will be acceptable. Is getting the endpoint a required step? I mean the objective is to run it as a webservice, so it is obvious that you will get it. I'm surprised it's even an explicit step here. Also, Training the model is required because you only created the model but never trained it. Starting the container is probably more important than getting the endpoint. Anyway... thus far, it looks like there is no consensus on the answer.

upvoted 2 times

KenCraw

4 years, 3 months ago

Any chance it could. be Train > Register > Start > Endpoint (as per the image)?

upvoted 1 times

...

allanm

4 years ago

You definitely need the endpoint if you are going to deploy anything as a web service. I don't think you need to train the model as it's already been trained. I think the right answer is - Create image > register image > start container > get endpoint

upvoted 1 times

...