Azure ML Online Endpoint Deployment Failing for HuggingFace Models - ResourceOperationFailure

Question

Hi Azure Community, I'm encountering persistent failures when trying to deploy HuggingFace models as online endpoints in Azure Machine Learning. The deployment fails during the provisioning stage with a ResourceOperationFailure. Error Message: ResourceOperationFailure: Internal error with reference to troubleshooting guide at https://aka.ms/oe-tsg#error-intern...
Steps I've tried:

1. Deploying through Azure Portal and SDK both
2. Using standard compute SKU (Standard_F4s_v2) What could be causing this internal ResourceOperationFailure?
   It is happening with all HuggingFace models and there is no log error message

Answer 1

I am not expert of this matter, but I have a doubt that your issue may be related to the subscription or the SKU.

Did you verify if your Azure subscription has sufficient quota for the resources you're trying to deploy ? Sometimes, the failure is due to exceeding the allocated quota for certain SKUs.

Try to follow this guide for troubleshooting : https://learn.microsoft.com/en-us/azure/machine-learning/how-to-troubleshoot-online-endpoints?view=azureml-api-2&tabs=cli#error-intern...

Also try to check that the Hugging Face model you're trying to deploy is compatible with the selected SKU (Standard_F4s_v2). Some models may require more memory or computational power than what this SKU provides.