DevOps for Machine Learning (Azure MLOps Part 3) - Train and Version Your Model

In this series of videos I'm showing how to get started with DevOps for Machine Learning (MLOps) on Microsoft Azure.

In the third video of this 5-part series, you'll discover how to setup Azure Machine Learning Services with your DevOps pipeline, as well as train and version your model.

And don't forget to click the bell so you don't miss anything.


📢 If you enjoyed this video help others enjoy it by adding captions in your native language -

🚀Recommended links to learn more about DevOps for Machine Learning (MLOps):

✅ Let's connect:

DISCLAIMER: This video and description contains affiliate links, which means that if you click on one of the product links, I’ll receive a small commission. This helps support my channel and allows me to continue making awesome videos like this. Thank you for the support!

#MLOps #DevOpsForMachineLearning #AzureML

[junkmail]

In Train Model step 5:58, before you move on, make sure Working Directory is set to training, otherwise the runtime won't be able to find the runconfig file.

[osafakarim]

It's seriously good. Awesome hands-on demo .

[andresurrea]

hi @Sascha Dittmann thanks for you learning in each video, but i have a question, the file train.runconfig is created automatically? or we have to create? thanks for your help

[pgdify]

Excellent work Sascha. Please keep coming up with new videos.

[RaviTejakarri]

Hi Sascha, In the Infrastructure creation there is an option to skip if workspace already exists. Is there any option for compute target/storage account? Like if it already exists skip the creation

[shaz-z]

Hi Sascha,
This is indeed a very good tutorial on MLOps so far I've seen, I would request you if you can create a video on how to keep a track of versioning dataset and model in azure, monitoring the model, how to trigger retraining pipeline when model performance start to deteriorate overtime and how to do A/B testing.

[yorjaggycastano]

Thks for ur dedication, I've found really useful your videos.
Very clear and well explained.

[sinabd]

Please make more videos in this topic. They're really helpful!

[joaquinaliagagonzalez]

Thanks you for the video !
If anyone can't find experiments site to see metrics and charts, inside the machine learning instance you have to launch the Azure Machine Learning Studio and there you'll find the metrics and charts seen in the video

[gtosXD]

Really helpful Sascha! Thank you again for this.

[ZulfiqarAli-xzy]

Hi, could you please explain where can I get my baseImage value for my train.runconfig file?

[saurabhverma]

Great work @Sascha . Thanks for making these videos mate. Really helpful and time-saving.

Wanted to understand one thing that If I manually create the “container registry” and allocate it with my machine learning workspace then it wouldn’t allow me to pull the docker images. Keep getting the “unauthorised user” error even though I am running with the owner access.

The same error doesn’t occur if I let the automated process to create ML workspace and its dependent services.

[Trendinglive]

Great series Sascha. 👏👏👏. Can you create a series for how did you created the git file. How to create all those install requirements etc. I believe it is done through vs. but it would be helpful if you can create a short video on how to create those folders and yml files

[xs]

Hello Sascha,
I am getting the following error continuously in Train model while running pipeline, can you please help me with this.
CondaValueError: Malformed version string '~': invalid character(s).
I attempted many things but still no luck.

[saurabhsupe]

Hi @Sacha, Thanks for adding a really detailed and easy to understand explanations.
I am facing an issue with access to container registry.
The error message is: "Unauthorized error encountered while attempting to connect to the container registry. Please try running resync keys on your workspace and retry.\".


Should I used a user based login in the script somewhere? Or could you point in the right direction to proceed.
Also could you also tell me that which part of the scripts create and access the ACR?

[arjunbutola]

Hi Sasha thank you so much for wonderful videos. previously it was working fine for me but when I ran pipeline again I am getting error while training "FileNotFoundError: [Errno 2] File does not exist: please help me on this

[temiwale]

Hey Sascha - I get 'worskpace' already exists when I re-run the pipelines. How do I bypass an existing AML ws?

[joaquinaliagagonzalez]

I have a question. What's the difference between requirements.txt and conda_dependencies.yml? Since you're installing almost the same libraries in both. Thanks!

[nirajkashyap]

Thank you very much for this video.. could you suggest me how i can integrate SAS Viya ML tool and make ci/cd in azure devops. Thank you

[bharathjc]

Unable to create resource manager connection says not an active subscription pls suggest