how to retrain model and deploy (if new model is better) by schedule or trigger with Azure MLops ?

Question

Hi

i am beginner of azure.

i am trying to use mlops.

it is not easy to not programmer...

but i want to practice mlops.

i want to retrain model by scheduling.

please let me know how to retrain model and deploy (if new model is better) by schedule or trigger with Azure MLops

Answer 1

Just a question, is there a way to automate the creation of a new web-service which contains the new model once retrain has be done. For example, I have created a retraining pipeline and I see that the model is performing better. Can I automate the process of deploying the new model behind an existing web-service or does this always have to be a manual job?

Answer 2

I'm interested in the above question as well, is there a way to accomplish this automatically without manually doing it?

Answer 3

Hi,

Once you have decided to retrain, you should:

Preprocess your data using a repeatable, automated process
Train your new model
Compare the outputs of your new model to those of your old model
Use predefined criteria to choose whether to replace your old model
A theme of the above steps is that your retraining should be automated, not ad hoc. Azure Machine Learning pipelines are a good answer for creating workflows relating to data preparation, training, validation, and deployment. Read Retrain models with Azure Machine Learning designer (preview) to see how pipelines and the Azure Machine Learning designer fit into a retraining scenario.

Please let me know if you have more questions.

Thanks,
Yutong

Answer 4

Hi,

AML pipeline is good way to help the model retraining with MLOps and Pyhton

This reference architecture shows how to implement a continuous integration (CI), continuous delivery (CD), and retraining pipeline for an AI application using Azure DevOps and Azure Machine Learning. The solution is built on the scikit-learn diabetes dataset but can be easily adapted for any AI scenario and other popular build systems such as Jenkins and Travis.

This is a sample of self-supervised learning with hyperparameter tuning and automated retraining: https://github.com/Microsoft/MLOps_VideoAnomalyDetection

MLOps doc: https://github.com/microsoft/MLOps
MLOps Python doc:https://learn.microsoft.com/en-us/azure/architecture/reference-architectures/ai/mlops-python
MLOps repo: https://github.com/microsoft/MLOps
MLOps pipeline: https://learn.microsoft.com/en-us/azure/architecture/reference-architectures/ai/mlops-python#mlops-pipeline