Platform X: AI Tips and Tricks

How do I use region restricted models?

Microsoft Azure deploys resources within regions. This specifies which of their server stacks your request is routed to. Often times the newest models are restricted to regions that recently have got new hardware. 

If a model has a region restriction, you must have a "foundry" type resource that is in that particular region. This region does not need to match the region of your virtual machine. If you have an existing foundry resource in another region submit a customer support ticket with the region you need to connect to. We will create another "foundry" type resource on your hub with that region.

Why am I not able to initially connect to Foundry?

If you are having trouble with initially connecting to Foundry double check that you have your "Local network access" set to "Allowed" within the browser you are using for AI Foundry. 

Why can I not connect to a model endpoint?

If you are getting a "Failed to create Azure OpenAI resource" error, try to deploy using the "Global Standard" mode instead of "Standard".

Why can I not deploy a model/project in Microsoft Foundry?

If you encounter a permissions issue deploying a model using a new project, you likely don't have permission to deploy a the new project. Currently, we only allow managers to create new projects. Please either use an existing project to deploy your new model or work with the lab bench PI to deploy a new project for you.