AI Studio - serverless endpoints - ServerlessModelNotAvailableInRegion

[harsimrit]

Hi,

I am trying to provision a serverless endpoint using azapi_resource for the azureml://registries/azureml/models/Phi-3.5-vision-instruct model in eastus2 but i am getting an error that "The requested model azureml://registries/azureml/models/Phi-3.5-vision-instruct is not available in region East US 2."

I am following the spec mentioned here -> https://learn.microsoft.com/en-us/azure/templates/microsoft.machinelearningservices/2024-04-01/workspaces/serverlessendpoints?pivots=deployment-language-terraform

Here is my code:

resource "azapi_resource" "ms_phi_35_vision_inst" {
  type      = "Microsoft.MachineLearningServices/workspaces/serverlessEndpoints@2024-04-01"
  name      = "phi-35-vision-inst"
  location  = "eastus2"
  parent_id = <id of the project>

  body = {
    sku = {
      name = "Consumption"
    }
    properties = {
      authMode = "Key"
      modelSettings = {
        modelId = "azureml://registries/azureml/models/Phi-3.5-vision-instruct"
      }
    }
  }
}

Complete Error:

│ {
│   "error": {
│     "code": "UserError",
│     "message": "The requested model azureml://registries/azureml/models/Phi-3.5-vision-instruct is not available in region East US 2.",
│     "details": [],
│     "additionalInfo": [
│       {
│         "type": "ComponentName",
│         "info": {
│           "value": "managementfrontend"
│         }
│       },
│       {
│         "type": "Correlation",
│         "info": {
│           "value": {
│             "operation": "df8877d5a1e63281627e84636220c4cf",
│             "request": "cca052e80839475b"
│           }
│         }
│       },
│       {
│         "type": "Environment",
│         "info": {
│           "value": "eastus2"
│         }
│       },
│       {
│         "type": "Location",
│         "info": {
│           "value": "eastus2"
│         }
│       },
│       {
│         "type": "Time",
│         "info": {
│           "value": "2024-11-05T18:57:50.6016318+00:00"
│         }
│       },
│       {
│         "type": "InnerError",
│         "info": {
│           "value": {
│             "code": "BadArgument",
│             "innerError": {
│               "code": "ServerlessModelNotAvailableInRegion",
│               "innerError": null
│             }
│           }
│         }
│       }
│     ]
│   }
│ }

I am able to to create a serverless deployment via the Studio UI and Azure CLI for the same model without any issue.

Any help/insights/guidance here.

[ms-henglu]

Hi @harsimrit ,

Thank you for taking time to report this issue!

From the error message, it seems that the model is not available in region East US 2. Would you please try to deploy it in another region?

Or if you could deploy this resource via Portal or other client tools, you could use the azapi vscode extension to help generate the configuration.

How to generate the azapi configuration with azapi vscode extension?

1. Copy the resource JSON. You could get the resource from Portal - JSON View, or az cli command like below:

az rest -m GET -u '/subscriptions/{subscription_id}/resourceGroups/{resource_group_name}/providers/Microsoft.MachineLearningServices/workspaces/{workspace_name}/serverlessEndpoints/{serverless_endpoint_name}?api-version=2024-04-01'

2. Paste the resource JSON to the terraform file, the extension will help convert it to the azapi config.

[florianchappaz]

Thanks for your input @ms-henglu.
I was facing the same issue and your command helped me to realize that I was missing the 'sku' block (and anyway the values inside are documented nowhere).