If it is impossible for GPT-4V to be used when it comes to images of human people, then, what is the next best vision model that can be used for interpreting those kind of pictures?
1 Like
Here is the solution i found:
The problem is the AzureOpenAI init args.
You should init it with
base_url=f"{api_base}/openai/deployments/{deployment_name}"
If you init AzureOpenAI not with base_url, but with azure_endpoint,
you will get the issue “Sorry, I can’t help with identifying or making assumptions about people in images.”
1 Like
its works for me, the base url in microsoft comes with base_url=f"{api_base}openai/deployments/{deployment_name}/extensions", when i delete “/extensions” its works