Add AnthropicLLM

[sdiazlor]

Closes #424

from distilabel.llm.anthropic import AnthropicLLM

if __name__ == "__main__":
    llm = AnthropicLLM(model="claude-3-opus-20240229", api_key="api.key")
    llm.load()
    output = llm.generate(
        inputs=[
            [
                {
                    "role": "user",
                    "content": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.",
                },
            ]
        ]
    )
    print(output)

[alvarobartt]

Hi @sdiazlor! Thanks for the PR, there's a couple of things I've mentioned and some others I'd like to confirm as:

• Is there any way we can check whether the provided model is an existing one? Could we offer a listing or just call an API endpoint to get the list of existing models within Anthropic?
• Could you reply with an example using the http_client? Or just mentioning which are the scenarios where it's useful? Also note that there may be issues when serializing the http_client if that appears to expect an httpx.Client instance

[sdiazlor]

• Is there any way we can check whether the provided model is an existing one? Could we offer a listing or just call an API endpoint to get the list of existing models within Anthropic?

@alvarobartt Thanks for your feedback.

1- We could do that by using something similar to this. Should we implement it?

from anthropic import AsyncAnthropic
def _check_model_exists(self) -> None:
    """Checks if the specified model exists in the available models."""
    annotation = get_type_hints(AsyncAnthropic().messages.create).get('model', None)
    models= [value for type_ in get_args(annotation) if get_origin(type_) is Literal for value in get_args(type_)]

    if self.model not in models:
        raise ValueError(f"Model {self.model} does not exist among available models.")

2- They use it to customize the httpx client as follows, so yeah, maybe this is an issue for serialization. Not sure how you handled this in similar cases, or if you just decided to remove it.

from anthropic import Anthropic

client = Anthropic(
    # Or use the `ANTHROPIC_BASE_URL` env var
    base_url="http://my.test.server.example.com:8083",
    http_client=httpx.Client(
        proxies="http://my.test.proxy.example.com",
        transport=httpx.HTTPTransport(local_address="0.0.0.0"),
    ),
)```

[gabrielmbmb]

Hi, it looks good!. Just need a small update before merging.