Search code examples
pythonazure-machine-learning-serviceazureml-python-sdk

Packaging multiple models from Azure ML experiment

So I have started to create a MLOps pipeline that is training multiple models within multiple pipeline steps.

The picture below is the graphical representation of the coded pipeline steps within the Azure ML Studio.

pipeline steps

These steps run fine and both the end steps produce the models I want (Train Data - Non EOW TFIDF and Train Data - EOW TFIDF)...

However this is where I get stuck with registering and packaging the model parts for deployment. These models get produced and stored within the individual pipeline step (see below for the model output of Train Data - Non EOW TFIDF)

non eow step output

but I don't know how I would register the model outputs from both pipeline steps together as the docs I have read for registering a model seem to only reference the ability to register one model from one path. https://learn.microsoft.com/en-us/python/api/azureml-core/azureml.core.model.model?view=azure-ml-py#register-workspace--model-path--model-name--tags-none--properties-none--description-none--datasets-none--model-framework-none--model-framework-version-none--child-paths-none--sample-input-dataset-none--sample-output-dataset-none--resource-configuration-none-

Basically, is it possible to produce multiple model outputs from multiple pipeline steps and register them together as one??

Thanks in advance for the help!

Solution

  • Here is sample for Multi-model Register and deploy. https://github.com/Azure/MachineLearningNotebooks/blob/master/how-to-use-azureml/deployment/deploy-multi-model/multi-model-register-and-deploy.ipynb