none
ML Studio Predictive Experiment will not deploy RRS feed

  • Question

  • Hi,

    I made an experiment which successfully trained a binary classifier on image data. Then, I clicked "Set up Web Service" which made a new predictive experiment which I ran successfully. But when I try to deploy the web service, the experiment fails and says an Unknown Error occurred. When I try to deploy it with the classic deployment, it says Service call failed. Error 500 (InternalServerError) when requesting /webservicegroups/8e76fa735d124c64b8cde4a416b117c7/modelpackages/

    Error Message: An unknown error occurred.
    Site Path: /webservices/deployExperiment
    Activity ID: dbe7d222-99a9-4a6b-98c0-4577aabb0cfa
    Request ID: 18ebd66b-05db-4be5-8874-8b905edcef26
    Workspace ID: 639d58d16d2b45fba5f4b2af7c0a5a70

    Workspace Type: PaidStandard
    User Role: Owner
    Tenant ID: f8cdef31-a31e-4b4a-93e4-5f571e91255a

    Tuesday, December 24, 2019 1:06 AM

Answers

  • I figured out what was wrong. In the training experiment, you need to put the saved model instead of making a model and training it. For all those looking at this post, doing "Set up web service" from the training experiment does not work correctly if you do it directly from the flowchart you made to train the model. If a lot of modules change into random stuff, you've done it wrong and will run into this issue.
    • Marked as answer by michael_46 Tuesday, December 24, 2019 2:59 PM
    Tuesday, December 24, 2019 2:59 PM

All replies

  • Hi Michael,

    Could you please try to re-deploy the web service as a new web service and check if it works? You can also post your experiment in the gallery and share the same so we can try to replicate the same and check if the error persists.

    Since this error is related to a single deployment we need to lookup the exact reason of this error for this request id. A support request might be needed so a azure support engineer can help you understand the reason for the failure if re-deploying the web service does not help. 

    -Rohit

    Tuesday, December 24, 2019 9:37 AM
    Moderator
  • I'm having trouble publishing the experiment to the gallery since the dataset I use is too large. Re-deploying the web service didn't help. But, if I go to web services, it says the web service was made but there are no endpoints.

    Tuesday, December 24, 2019 1:00 PM
  • I figured out what was wrong. In the training experiment, you need to put the saved model instead of making a model and training it. For all those looking at this post, doing "Set up web service" from the training experiment does not work correctly if you do it directly from the flowchart you made to train the model. If a lot of modules change into random stuff, you've done it wrong and will run into this issue.
    • Marked as answer by michael_46 Tuesday, December 24, 2019 2:59 PM
    Tuesday, December 24, 2019 2:59 PM
  • Hello,

    Setting up the web service from the training experiment helps to setup a retraining model. The steps to create and use the same is available here.

    Using the predictive experiment for deploying the web service is the correct way to setup the service which can be updated with a new trained model from the retraining experiment.

    -Rohit

    Thursday, January 2, 2020 6:26 AM
    Moderator