Accessing different model versions from same endpoint

Question

I am using model versioning and would like to have different model versions accessible via the same endpoint. Any best practices to access the multiple models from the same endpoint.

Answer 1

@AdityaSa Thanks, You may deploy “locally” to a Azure Machine Learning compute instance, by specifying different port # for each version. They are converted to a URL according to the format https://<compute instance’s name>-port.region.instances.azureml.ms/score

Model v1: service_url = https://azure-ml-compute-instance-name-8001.westeurope.instances.azureml.ms/score
Model v2: service_url = https:// azure-ml-compute-instance-name-8002.westeurope.instances.azureml.ms/score

There’s sample code in documentation. You can specify port to deploy with the following parameter.
deployment_config = LocalWebservice.deploy_configuration(port=8001)

We recommend using the new ML Endpoints (Preview) What are endpoints (preview) - Azure Machine Learning | Microsoft Learn.