Forum Discussion

teeceegee's avatar
teeceegee
Regular Visitor
2 years ago
Solved

Error when trying to use copilot

We have been experimenting with Fabric for a little while now while we examine whether it is appropriate for us to use as a data platform. We turned our paid premuim instance from an F16 to an F64 and got access to the "copilot" button in the ribbon. However, when we ask a question of copilot we get the following error messge:

An error occurred while parsing the MWC token response: b'{"Message":"F16 SKU Not Supported","Source":"ML","error_code":"PERMISSION_DENIED"}' An error occurred while generating the MWC token: An error occurred while parsing the MWC token response. Retrying in 0.1 seconds...

 

We are definitely on an F64 instance. Is anyone else experiencing this or have any suggestions as to how to fix?

  • AndyDDC's avatar
    AndyDDC
    2 years ago

    OK so the Azure SKU was set to F16, you've then scaled it to F64 and still having issues.

     

    Can you try the following:

    • Pause and resume the capacity while it's set to F64
    • Create another Workspace and assign the F64 capacity to it
    • Ensure that both the toggle switches are on for "Copilot and Azure OpenAI Service (preview)​" in the Admin portal
    • Create a new Notebook in the new workspace, click the Co-Pilot button and test again

5 Replies

    • AndyDDC's avatar
      AndyDDC
      Most Valuable Professional

      OK so the Azure SKU was set to F16, you've then scaled it to F64 and still having issues.

       

      Can you try the following:

      • Pause and resume the capacity while it's set to F64
      • Create another Workspace and assign the F64 capacity to it
      • Ensure that both the toggle switches are on for "Copilot and Azure OpenAI Service (preview)​" in the Admin portal
      • Create a new Notebook in the new workspace, click the Co-Pilot button and test again
      • teeceegee's avatar
        teeceegee
        Regular Visitor

        Hello,

         

        Apologies for taking a little while to respond. Due to other commitments it was a couple of days before we could try this out.

         

        Pausing and resuming the capacity appears to have resolved the issue. Thanks for the work-around and we assume that at this time this is a bug.