0

I am currently trying out different architectures with Azure ML Ecosystem. Currently, I am testing out Azure ML Studio Designer.

I want to create a complete End to End ML system, where I train several models and deploy the best.

The Pipeline Created: enter image description here

In the Designer, is it possible to register a custom-trained model(it will appear in the model assets in the workspace)? Also is it possible to directly deploy the best model within the Designer?

Not manually clicking on the job and Registering and Deploying but in a more automated way.

Improve this question

1 Answer 1

Reset to default
1

First, we need to create the designer manually and assign the deployment manually for the first time. Input for the next iteration will be the web service input. For that we need to get the web service output to be attached with the evaluation model metrics. For the next time it will be running in repeated state.

Follow the steps to make the automation for the case of designer deployment from the designer itself.

enter image description here

enter image description here

enter image description here

The above images are not being connected to the web service input for the automation.

enter image description here

The above image will be taking the result of automation of the web service to perform the continuous deployment.

Connect to the compute target.

enter image description here

enter image description here

Create new pipeline job

To deploy we need to have an inference cluster. It will be AKS by default.

enter image description here

enter image description here

enter image description here

enter image description here

We will get the inference pipeline like above. Get the details of the web service endpoint and deploy it with the web service. First time have to do it manually, next time it will work automatically.

enter image description here

Above is the total structure will look like in designer for automation of the deployment.

Improve this answer
Sign up to request clarification or add additional context in comments.

4 Comments

By doing so are the Models versioned and shown under Models Assets in the studio?
Versions will be shown. For every execution. But it can get manually. not automatically.
When I try to generate the Inference Pipeline it says I need to select only one model. Does that mean in the designer we cant select the best model and deploy it automatically, is it?
For example I gave linear regression as one model. Multimodel classification options need to be replaced in the design. It will function accordingly. then we can get the different results and then can get the best model. but every result we get will be deployed automatically. not only the best model. If we need the best model, we need to check the evaluation metrics of the models and take MAX operation, then separate the required model from the list and deploy. It's too complex to execute. But the success rate is low.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.