Last time I was figuring out how to actually use the ML Model Endpoints in Fabric: https://www.reddit.com/r/MicrosoftFabric/comments/1qtuicy/ml_model_endpoints_anyone_used_them/
I've been investigating this feature a bit when I've had time to and found a pretty annoying bug (or feature?):
If I pause a capacity with some ML Model endpoint activated the said endpoint will go to a state that after capacity restart It won't go live anymore. I need to re-create the Model and activate the endpoint again to get it back alive. When it goes to this state the deactivation does not work also.
Has anyone else stumbled upon this?
I'm restarting the capacity because this is a demo that is needed from time to time. :)
Steps to reproduce:
Create and publish ML Model
Activate ML Model endpoint with "Activate version endpoint"
Restart Fabric capacity
Try to use the endpoint for example with the "Preview predictions" feature