Hope to cancel "required_args" mecanism. Upgraded openai-python (v0->v1) refuses argumment "repetition_penalty"

[TissueC]

Confirm this is an issue with the Python library and not an underlying OpenAI API

• This is an issue with the Python library

Describe the bug

I notice that about 2 weeks ago vllm has merged pr about adding an argument repetition_penalty aligned with huggingface. (refer to pr 1424)

However, openai-python also graded to v1 several days ago. it refuses the unknown argument repetition_penalty. The following is the detail error.
xxx/examples/openai_chatcompletion_client.py", line 72, in <module> chat_completion = client.chat.completions.create( File "/opt/conda/lib/python3.10/site-packages/openai/_utils/_utils.py", line 299, in wrapper return func(*args, **kwargs) TypeError: Completions.create() got an unexpected keyword argument 'repetition_penalty'

This conflict makes that we can not depend openai library for open-source models and self-defined arguments (much more than repetition_penalty). This library has been factually used in many API-based model deployment library like fastchat and vllm.

I really hope that the required_args can be removed. At least, we can modify its behaviors without modifying the source code of openai-python.

To Reproduce

1. git clone vllm and setup it.
2. Make some request using vllm and openai, add arguments repetition_penalty
3. The error happens.

Code snippets

No response

OS

Linux

Python version

Python v3.10

Library version

openai v1.2.2

[rattrayalex]

Thanks for reporting, we'll consider this. We us that to help enable good typing of overloads, which is an important feature.

[RobertCraigie]

This is not anything to do with required_args, what you're trying to do is pass a custom request parameter which is supported by the extra_body argument.

from openai import OpenAI

client = OpenAI()
client.chat.completions.create(
  model='gpt-4',
  messages=....,
  extra_body={'repetition_penalty': 2},
)