none
Major Version Upgrade on 12/8/14 May Require User Action RRS feed

  • Question

  • Release Update: May Require User Action

    The December 8, 2014 release of Azure Machine Learning includes a major version upgrade of modules and machine learning framework. The upgrade will have following impact on your models:

    • Any saved and trained models created before upgrade will become unusable in experiments.
      • Action: You must re-run the training experiment and save the trained model again. Then go to corresponding scoring experiment and upgrade the saved trained model. For more information on how to perform these steps, see this forum article.
    • When you open an existing experiment, you will be prompted to upgrade modules in your experiment.
      • Note that for large experiment, the upgrade may take few minutes.
      • You can view the previous version of the experiment by clicking “View Run History”. However, you will be unable to edit or run the previous versions.
    • Any new experiments will automatically use the new versions of modules. No action is required.
    • Already published web services are unaffected by the upgrade and continue using the previous version of modules. No action is required.
    Tuesday, December 9, 2014 6:10 PM

Answers

  • Hello:

    Re-running my models will take at least a week of compute time.  In the meantime I need to do some scoring.  Is there a way I can recover a former run with existing models and scoring module so I can get my results?

    Thanks for your help.  This is the experiment (which now fails):

    https://studio.azureml.net/Home/ViewWorkspace/ffac4490b5e448fa8778ceadc74a0b97?&hashSeg=#Workspaces/Experiments/Experiment/ffac4490b5e448fa8778ceadc74a0b97.f-id.45c727a9f5654d5f9a9b2629d25c0669/ViewExperiment

    • Proposed as answer by Ga_G2 Friday, April 3, 2015 11:28 AM
    • Marked as answer by neerajkh_MSFTModerator Saturday, April 25, 2015 6:27 PM
    Sunday, January 18, 2015 3:43 AM

All replies

  • Note also that Score Recommender module returns recommendations for same user in a single row, as opposed to multiple rows.
    Monday, January 12, 2015 5:55 PM
    Moderator
  • Hello:

    Re-running my models will take at least a week of compute time.  In the meantime I need to do some scoring.  Is there a way I can recover a former run with existing models and scoring module so I can get my results?

    Thanks for your help.  This is the experiment (which now fails):

    https://studio.azureml.net/Home/ViewWorkspace/ffac4490b5e448fa8778ceadc74a0b97?&hashSeg=#Workspaces/Experiments/Experiment/ffac4490b5e448fa8778ceadc74a0b97.f-id.45c727a9f5654d5f9a9b2629d25c0669/ViewExperiment

    • Proposed as answer by Ga_G2 Friday, April 3, 2015 11:28 AM
    • Marked as answer by neerajkh_MSFTModerator Saturday, April 25, 2015 6:27 PM
    Sunday, January 18, 2015 3:43 AM