none
ML Server integration with Azure ML Services RRS feed

  • Question

  • Hello,

    is ML Server completely independent from Azure ML Services (Workbench, Model Management, Experimentation Service, etc) and vice versa?

    Are there plans to integrate these products and services, for example:

    - build models using ML Server libs (revoscale, microsoftml) from Workbench

    - manage them with Model Management

    - deploy them via ML Server operationalization (not like service of ACS)

    - manage and deploy models built with ML Server libs via Azure ML Services (using Model Management and ACSs)?

    Thanks

    Sunday, February 25, 2018 10:16 PM

All replies

  • Hi,

    The high performance ML packages from ML Server can work seamlessly with AML Services (workbench, model management and experimentation service). Please refer to the samples https://github.com/Azure/MachineLearningSamples-mml  on how to use them with AML services.

    In brief, we offered those ML Server packages such as RevoScalePy, MicrosoftML as standalone Python packages that can be downloaded (PIP installed with a dedicated URL) and used as any of other Python packages in AML services, for experimentation, model management, and deployed to ACS. On the other hand, you can also use those packages to do training on AML environment, and use the azureml-model-management-sdk to deploy via ML Server operationalization.

    I am the PM working on this integration. Look forward to working with you more closely. Hope Microsoft ML can benefit your business.

    Thanks,

    Carl

    Monday, February 26, 2018 1:03 AM
  • Hello,

    thanks for your answer. Will keep you informed about our experience.


    • Edited by vpolezhaev Monday, February 26, 2018 12:51 PM
    Monday, February 26, 2018 12:50 PM