Power BI is turning 10! Tune in for a special live episode on July 24 with behind-the-scenes stories, product evolution highlights, and a sneak peek at what’s in store for the future.
Save the dateEnhance your career with this limited time 50% discount on Fabric and Power BI exams. Ends August 31st. Request your voucher.
I have registered my ML model in fabric (free capacity) like in the image below. There are 2 versions of it, and i tried to activate the version endpoint so that I can call it in KQL eventhouse for streaming prediction purpose. But i do not know why i can not activate either version 1 or 2, it always displays failed status. Can anyone have experience on this help me ? Thank you so much.
Note: i have enabled the endpoint ML in Admin portal already, and i can use notebook to call the model, but it will take time to load model so that why i do not use notebook to load and use endpoint instead.
Solved! Go to Solution.
hi, i found the problem my own, i trained the spark model and save in mlflow but fabric have not support endpoint for this so i can not activate it. Thanks all
hi, i found the problem my own, i trained the spark model and save in mlflow but fabric have not support endpoint for this so i can not activate it. Thanks all
Hi @bao_phan ,
This usually happens on Fabric Free Capacity because the ML endpoint feature has some limitations (resource, scaling, or admin restrictions). A few things to check:
Fabric Free Capacity: ML endpoints may not be fully supported or have limited availability on free workspaces. Upgrading to a paid capacity (F64 or higher) often resolves this.
Admin Portal Settings: Double-check that “ML endpoints” is enabled for your workspace and user permissions are correct.
Model Requirements: Ensure your model files (dependencies, environment YAMLs, etc.) are complete and compatible.
Logs: Look for detailed error logs or messages (sometimes available in the workspace activity or endpoint monitoring) to pinpoint the exact issue.
Since you can use the notebook to load the model, it’s likely a resource or permission limitation on the endpoint side.
Tip: Try deploying in a Pro or Premium workspace if possible.
is it possible to call the endpoint inside KQL or do i have to load the model there, i just read this article and i did the load model option: https://blog.fabric.microsoft.com/en-US/blog/ml-model-scoring-in-fabric-eventhouse-via-update-policy...