Support for GPT-4 and 0613

[MiguelElGallo]

This issue is for a: (mark with an x)

- [ ] bug report -> please search issues before submitting
- [X] feature request
- [ ] documentation issue or request
- [ ] regression (a behavior that used to work and stopped in a new release)

Minimal steps to reproduce

In this blog post, says that this repo will use GPT-4 models:
https://techcommunity.microsoft.com/t5/ai-cognitive-services-blog/working-with-gpt-4-and-chatgpt-models-on-azure-preview/ba-p/3773595

Any log messages given by the failure

I cannot find anywhere that this REPO is deploying GPT-4 models for Azure OpenAI.

Expected/desired behavior

Detail in the documentation or update a parameter to switch to use GPT-4

[vrajroutu]

Yeah!! Looks like there is some rework required here

[AllaeddineD]

Experiencing the same issue. Did you manage to make it work ?

[ThomasLambertES]

same here, got preview access to GPT-4 and i'm looking into deploying in that model.

[sk1136]

Any update on this? Can we use this same code with GPT-4 and embeddings. The interface for GPT 3.5 and 4 is different so we can't just change the model name and version number.

[vrajroutu]

I'm using GPT-4-32k the whole chatreadretrive code should be changed. I will push the code soon. But note that response will be slower when compared to Gpt-3.5-turbo but quality is really good.

[ThomasLambertES]

I'm using GPT-4-32k the whole chatreadretrive code should be changed. I will push the code soon. But note that response will be slower when compared to Gpt-3.5-turbo but quality is really good.

please do tell us once it's ready, i'm really looking forward to test

[rabbyn]

Hi @kishorerv93 , can we now use GPT-4 (or 4-32k) ? Thanks

[PTTrazavi]

According to (Azure OpenAI), gpt-35-turbo (version 0613) is available now. Same as gpt-4 (version 0613), they only work with openai.ChatCompletion.create, but not openai.Completion.create.
Can someone modify the chatreadretrieveread.py code so it can work with openai.ChatCompletion.create? Thanks!

[pamelafox]

A PR is in progress: #385

[vrajroutu]

@AllaeddineD @PTTrazavi @rabbyn Submitted my PR for GPT-4 #394

[pamelafox]

This repo got ported to support Chat Completions a while ago (0613). It defaults to 3.5 however, since most don't have access to 4. To use GPT-4 :

In infra/main.bicep, change chatGptModelName to 'gpt-4' instead of 'gpt-35-turbo'. You may also need to adjust the capacity above that line depending on how much TPM your account is allowed.