Migrating Projects from Legacy Engines

You can migrate legacy engine projects to use ML Runtimes.

This example uses the AMP Churn Modeling with scikit-learn AMP to guide you through migrating a project to ML Runtimes.
You can start from any Project that has a Legacy Engine configured and have Jobs scheduled or Models and Applications deployed.
  1. Open the selected project that uses the Legacy Engine.
    Observe the current Project Settings options.
  2. Using the Project Settings page, switch to ML Runtime.
    You should be able to see the ML Runtime selector when you start a new Session.
    The Project Overview page might display warnings for workloads (for example, Jobs) that are configured to use Legacy Engines.
  3. If the project has jobs, update each job to use your preferred ML Runtime.
    In this Churn modeling example, switch each workload to the following:
    • Editor - Workbench
    • Kernel - Python 3.6
    • Edition - Standard
    • Version: 2021.02 Runtimes
    1. You can now use the following sequence to retrain your Churn model using ML Runtimes via the updated jobs:
    • Install dependencies.
    • Ingest data.
    • Train Model.
  4. If the project has models, redeploy each model to use ML Runtimes.
    Validate that the endpoint is using ML Runtime.
    You can also test the Model.
  5. If the project has applications, redeploy each application to use ML Runtime.
    The Applications detail page will display a warning that the Legacy Engine is configured.
    1. Redeploy the application based on the required ML Runtime.
    2. When the application is restarted, check to ensure it is up and running.
  6. Optionally, you can follow the same steps to update Runtime kernels from using Python 3.6 to Python 3.7.