question

27051995 avatar image
0 Votes"
27051995 asked ·

update real interference pipeline

I deployed my Training pipeline and my Real-time inference pipeline.
With the REST-Api of my training pipeline I'm able to retrain my ML model. Is it possible to use that retrained model automated in my real inference pipeline?
When i trigger the pipeline in ML studio I have to update my real inference pipeline manually. Since I want to trigger my retraining external that is not possible.
Thanks in advance.

azure-machine-learningazure-machine-learning-studio-classicazure-machine-learning-inference
10 |1000 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

GiftA-MSFT avatar image
0 Votes"
GiftA-MSFT answered ·

Hi, here's a reference on which technology to use based on a given scenario. For your scenario, you should be able to create an Azure Machine Learning pipeline using the SDK to trigger a pipeline based on a time/change based schedule and then update the web service accordingly. Depending on the complexity of your triggers or data prep needs, you can leverage other technologies such as Logic Apps or Azure Data Factory to trigger your Azure Machine Learning pipeline. Currently, you can only use the Azure Machine Learning SDK to automatically update the web service. Hope this helps.


·
10 |1000 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

GiftA-MSFT avatar image
0 Votes"
GiftA-MSFT answered ·

Hi, when you deploy a web service from designer, you can select to "deploy as a new real-time endpoint" or "replace an existing real-time endpoint". The option (replace an existing real-time endpoint) enables you to update the previous endpoint. Currently, there's no way to trigger an update to real-time endpoint, it's still a manual process (I will verify and share updates). To programmatically update your web service using Azure ML SDK, please refer to this document.

· 1 ·
10 |1000 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

Hi @GiftA-MSFT ,
thanks for your answer. I found this link were they say it is possible to update my endpoints via Data Factory. Unfortunatelly this is not described well. Is there a new updated version of this blog entry?
My goal is to update my endpoint once in a month because I use "supervised learning" and I can get new data. For that I want to upload or export my dataset to my blob storage and update my ML model and my endpoint automated. Is this possible with this approach?

Thanks!


0 Votes 0 ·