gpt-4o
ChingLuoH opened this issue · comments
Is your feature request related to this sample app, or to an Azure service, such as Azure OpenAI or Azure AI Search?
Please make this app compatible with gpt-4o.
Is your feature request related to a problem? Please describe.
Getting error: OpenAI.APIStatusError: Request Header Fields Too Large when testing with gpt-4o model. The app works fine when using gpt-35-turbo-16k, gpt-4, gpt-4-turbo.
Describe the solution you'd like
This app support gpt-4o as a LLM model option.
Is this feature specific to your use case or your organization, or would it apply broadly across other uses of this app?
It applies to all.
Describe alternatives you've considered
Additional context
im using gpt4.o flawlessly no issues
Hi @ChingLuoH , I am not able to repro this using my AOAI resource in EastUS2, but just wanted to ask if you can share which region you're using for AOAI? I can check to see if I can repro it in your region.
It may also help if you can share the full stack trace, to verify if the issue is with the webapp or in the underlying service call. Based on what you've shared above, it looks like it may be coming back from the API, in which case the problem is not with the webapp code but with something else.
I am seeing this in East US, trying East US 2 next
My Azure OpenAI is in East US region having this error.
I'm seeing the same HTTP 431 errors in East US 2. I will try next with a gpt-4-32k model instead.
Working great in East US using gpt-4-32k. No quota for that model in East US 2 so couldn't test. I'm also using vector search on my data. The problem seems to be something new with the gpt-4o model, maybe other features causes it to exceed header length when making calls? I've got some more deployments to do tomorrow and will try some different combinations.
The same issue is also happening in the Chat playground when using your own data with a gpt-4o model. Switching to a gpt-4 model with no other changes makes it work, so it must be something with the gpt-4o model.
I am also seeing this error in East US 2
Hi everyone, I got confirmation from the service team that this is a known issue impacting GPT-4o models. There is a planned mitigation which will begin rolling out next Tuesday after the holiday, so by EOW next week we should see this resolved for all impacted regions.
I'll follow up here with more information as it becomes available.
Hello!
We are having the very same issue today with EastUS and EastUS2 regions, using gpt-4o with BYOD and Azure AI Search.
Do you have any ETA when a fix will be available?
Thank you.
We're experiencing the same behavior in EastUS, EastUS2 y NorthCentralUS regions. We detected this issue when using the integration with Azure AI Search.
Same issue here - work good on Azure AI studio , getting Error Request Header Fields Too Large on the deployed web app.
We have an OpenAI balancer by APIM and some consumers thought that it was an APIM problem. Any region without that issue? Confirmmed for East US, East US 2 and North Central US. I will try it on the others disponible regions tomorrow
FYI... Azure AI Studio Chat Playground started working over the weekend when using your own data.
The same issue is also happening in the Chat playground when using your own data with a gpt-4o model. Switching to a gpt-4 model with no other changes makes it work, so it must be something with the gpt-4o model.
So, is it working? Is it fixed?
I am still getting the error in east us 2.
I'm still getting the error in our web app deployed from the azure ai studio but not getting the error from within the chat playground.
I am also getting the same issue even after upgrading the openai version to 1.30.5
#898
It seems to be fixed this morning when I tried. I made no changes.
It is definitely fixed in AI studio but not sure it's fixed in the webapp
I am still seeing the same error with latest code:
Maybe the difference now is the fact that I'm using chat history and you are?
Yes - chat history is enabled
It is working for me now. If it is not working, are you using API version 2024-05-01-preview?
GPT-4o is working now in East US region.