Deploying a Model with the Azure ML Designer

Рет қаралды 7,286

Күн бұрын

In this video, we will deploy a machine learning model to an Azure Container Instance for "real-time" inference using the Azure ML Designer's drag-and-drop user interface.
LINKS AND INFO
Catallaxy Services - www.catallaxys...
Chicago parking tickets dataset -- cspolybasepubl...
Chicago parking tickets database -- sqlsunday.com/...

Пікірлер: 33

@rickpack6700 2 ай бұрын

Taking DP100 today and found your video after scoring the lowest on “Deploy and retrain a model”. Thank you!

@KevinFeasel 2 ай бұрын

Good luck on the exam, Rick.

@rickpack6700 2 ай бұрын

I passed! These videos helped. Thank you for putting so much effort into these, Kevin.

@SuperChrisDub 6 ай бұрын

I laughed in recognition at your "pedantry in the comments" joke. I recognize that trait in myself and in many of my colleagues over the years. That pedantry is part of the programming/data analyst mindset. we need to make these kinds of fine distinctions to be good at our jobs; it just spills over into our world-view more generally! From, A nerd - and proud of it Class of Sinclair ZX81 (with 16KB ram pack, if memory serves me correctly) (No more are we silo'ed off in the basement, called bearded weirdies by our colleagues or mocked for our social skills. The game has changed. We hold the keys to the building now!)

@Michael-lb3hh 7 ай бұрын

I love this series. Such an amazing find. Keep it up! Your teaching is fantastic

@VOGTLANDOUTDOORS 6 ай бұрын

A WONDERFULLY-DETAILED, FUN and TIMELY VIDEO - nicely done ! - Mark in North Aurora IL

@user-th2vd4yr9v 10 ай бұрын

kevin i am really enjoying and appreciating your content and quality and style - it is helping me transfer 30 years of AI experience over into Azure ML which is still new to me. I had a comment about defining one-off vs. batch inferencing, and differentiating between real-time and queued processing, but it's clear Microsoft has high-jacked and blurred these terms in AML. Typical. *sigh* keep up the great work. Dr. Vogt

@scottsilvasy7855 3 ай бұрын

👊 Well done Kevin. Thanks - one of the better tutorials I have ever seen.

@simulacrum443 4 ай бұрын

Super useful video and the best tutorial playlist on Azure ML 💯

@ivancasanagallen 2 ай бұрын

Excellent work, Kevin! You make complex concepts easy to digest. I was wondering if you could share how to deploy a model generated with Auto ML. A Voting Ensemble, for example. My guess is that Voting Ensemble is going to be the best shot for many users, at least initially. That´s why I am thinking it would be interesting to see how to produce a batch endpoint. Thanks!

@KevinFeasel 2 ай бұрын

AutoML model deployment is something I can add to my backlog, sure. It may be a bit before I'm able to get the video out, so I'd recommend checking out the docs on this in the meantime: learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-automl-endpoint?view=azureml-api-2&tabs=Studio for deploying an AutoML endpoint (online, not batch) and learn.microsoft.com/en-us/azure/machine-learning/how-to-use-batch-model-deployments?view=azureml-api-2 for batch deployment.

@caiyu538 7 ай бұрын

great lectures, great teacher

@bigworldparty 5 ай бұрын

Thanks for the video series. Something I'm not clear about, why publish (meaning make an endpoint for usage) a pipeline rather than just a model? When I publish the pipeline, then I must provide a bunch of col values that may not be used by the model (if the pipeline "Select Columns In Dataset" has deselected some cols). On the other hand, if I were able to publish only the model (as we do with the AutoML), then I would have less overhead and only need to pass the values the model actually requires. Also I like how Azure lists the AutoML models under the "Models" menu, compared to the pipeline related models which are not listed there.

@KevinFeasel 5 ай бұрын

The main reason to publish the pipeline is so you can apply the same transformation rules to production data that you did for your training data. This becomes more relevant when you get to code-first model development in Azure ML, where you can get a lot more sophisticated around how you're handling the data. If you have features that the model does not use at all--that is, they are not columns on the DataFrame you use in training--you can modify the shape of the input, like I did to remove the PaymentIsOutstanding column from the input, starting at 5'33".

@tianhaoluo6782 2 ай бұрын

Thanks for the wonderful series! One thing I don't quite understand: why do we need the Enter Data Manually and Edit Metadata parts? Wouldn't it suffice to use Web Service Input only since it is an inference job? Does it basically 'tell' the model that now we do not have the labels anymore?

@KevinFeasel 2 ай бұрын

You're exactly correct. If we stop right after adding the Web Service Input, the model we deploy will work fine, but our web service will expect exactly the same input columns as our initial CSV file. That includes the PaymentIsOutstanding label column, which would be silly to include in production. For that reason, we remove the ChicagoParkingTickets file input and substitute it with a manual version that excludes PaymentIsOutstanding from our expected inputs.

@ashimapahwa778 7 ай бұрын

Really appreciate your content, you make it so easy to deploy the model. Would you have any advice on consuming this model in Tableau? General advice is to use a Web data connector but i am bit lost

@KevinFeasel 7 ай бұрын

Hmm, unfortunately, I'm not that familiar with Tableau but my presumption is that the web data connector has you call out to some URL. In this case, that would be the URL of your Azure ML service's score endpoint. For example, with Azure Container Instances, it might be {some GUID}.{region}.azurecontainer.io/score. Your body would be in the shape of the dataset, so something like { "data": [ { "input1": 123, "input2": 456, ... } ] }, similar to my Postman request in the video. You may also need to include a bearer token, depending on how your service is secured. Hopefully this helps you out.

@hinaque4505 2 ай бұрын

Thank you for these amazing videos, can you please help me understand what is postman? and is it part of the Azure studio, Thanks A LOT!

@KevinFeasel 2 ай бұрын

Postman (www.postman.com/) is not part of any Azure or Microsoft product. It is a tool you can use to test REST APIs. I use it in this video (and in quite a few other places) because it provides me an easy way of making an API call, seeing the response, and even building tests around how I expect the response to look.

@hinaque4505 2 ай бұрын

@@KevinFeasel Thank you so much for your prompt reply, your videos are amazing!!!

@113because 2 ай бұрын

Thạnk your video. But, i did yourstep suggestion , but I got error "azureml.studio.common.error.ColumnNotFoundError: Column with name or index "Label" does not exist in "Dataset", but exists in "Transformation". Please to help me?

@KevinFeasel 2 ай бұрын

If you're following along with the videos in this series, we don't have a Label column at all. For that reason, I might not be able to tell you for sure why you're getting that error. My best guess, given current information, is that perhaps you have something in the transformations on the training side, renaming the PaymentIsOutstanding column as Label. If you did that, then your inference would expect the PaymentIsOutstanding column because it would need something to rename as Label.

@113because 2 ай бұрын

Thanks yours responding, Label is a column in my data which used to make a pipeline. First i used label is my target for training, after that i want to deploy a endpoint model, which can inference with new data (without label/target), but I fail. Do you have a suggestion?

@KevinFeasel 2 ай бұрын

@@113because Given current information, as I noted above, you might be using the Label column in one of your transformations (column renames, data type changes, cleaning missing data, etc.--basically, everything other than loading data, training the model, scoring the model, and evaluating the model). If so, then you'd have to remove Label from those transformations in the training section; otherwise, scoring will require Label in the Apply Transformation step.

@zubairmushtaq7912 10 ай бұрын

kevein please make a video on batch inference

@KevinFeasel 10 ай бұрын

The code-first ML series uses batch inference, but thanks for the suggestion. I'll show how to do this in the AML designer in a future video as well.

@zubairmushtaq7912 10 ай бұрын

@@KevinFeasel yes, please make a video

@shaikathaharnawaz1718 8 ай бұрын

@kevin designer video seems missing in the playlist , can you please upload

@KevinFeasel 8 ай бұрын

Which designer video do you mean? The two videos I did on the Azure ML Studio Designer are in the AML playlist (kzfaq.info/sun/PLeWL8zChJ2uuGs-8nvxMdo26XkYGWt7ui). This video is the first and the second is Batch Processing in the Azure ML Designer (kzfaq.info/get/bejne/bpOWlptq3pjMZ6c.htmlsi=54Vjubq3ps1Qka3P), which is the latest video in the playlist as of today.

@shaikathaharnawaz1718 7 ай бұрын

@@KevinFeasel I was going through the playlist kzfaq.info/get/bejne/j7KTlqt5nMi3Z5s.html and the third video seems to be broken, instead of deploying model you can add training a model with designer and it will complete :). just FYI

@KevinFeasel 7 ай бұрын

@@shaikathaharnawaz1718 You are absolutely right. I neglected to add that video (kzfaq.info/get/bejne/frWapruBs52th6c.html) to the playlist. I've fixed that mistake, so thank you for bringing it up.

@shaikathaharnawaz1718 7 ай бұрын

@@KevinFeasel Thank you for the work...it's really helping me a lot, especially SDK v2 very few online contents out there