BootstrapFewShot failing due to lm.copy for AzureOpenAI

[insop]

Since #521 is closed state, creating a new issue.

Now, the initial PR #522 for the issue was reverted by this PR, #584 recently, I think we need to keep this issue.

---

BootstrapFewShot failing due to lm.copy for AzureOpenAI, likely due to positional argument issue from recent changes.

• Issue: BootstrapFewShot failing due to lm.copy for AzureOpenAI #521

Environment

• code: 98304a2
• model azure openai

Error message

• during BootstrapFewShot run, code errored out due to missing arguments, api_version

>[dspy/dsp/modules/lm.py](https://file+.vscode-resource.vscode-cdn.net/Users/insop/Projects/Github/NLP/dspy/dsp/modules/lm.py)(106)copy()
    102         model = kwargs.pop('model')
    105
--> 106         return self.__class__(model, **kwargs)

Changes

AzureOpenAI expects positional arguments, so it is failing. I am putting my change for review if that will impact other models.

class AzureOpenAI(LM):
...

    def __init__(
        self,
        api_base: str,
        api_version: str,
        model: str = "gpt-3.5-turbo-instruct",
        api_key: Optional[str] = None,
        model_type: Literal["chat", "text"] = "chat",
        **kwargs,
    ):

[arnavsinghvi11]

thanks for circling back on this @insop . What were the proposed changes for handling this?

[insop]

Thank you @arnavsinghvi11 .

I catch up the latest main and will propose soon.

[snimu]

I've run into the same issue, and it looks to me like the problem isn't8 that api_base needs to be renamed to azure_endpoint, but that dspy.AzureOpenAI hasn't redefined the LM.copy-method, so it falls back on that. For that method to work, self.kwargs in AzureOpenAI would have to include api_base and api_version, but they don't. So either, AzureOpenAI.copy should be implemented (preferred), or api_base and api_version have to be added to AzureOpenAI.kwargs.