Robin Picard
Robin Picard
Looks good overall, but I would not fetch images from an url for local models as it's adding troubles for us. Also I would avoid checking "image.data.startswith(("http://", "https://")):" in the...
Yes I wanted to do that initially, but I'm worried it would affect some users that expect the `eos_token` not be accepted in their current implementation. For example, there's the...
Yes, I closed the PR a few minutes ago as I also realized that it's not a viable option.
Awesome! Don't hesitate to reach out if you have questions.
I'm not sure I follow you. To me the issue is a lot more straightforward: we do not generate the correct regex for the output type `List[Literal["Paris", "London", "Rome", "Berlin"]]`...
i looked more into it and it's really quite tricky. We should probably solve that in the context of a wider refactoring of the output type system.
Hi @ahmeda335! We modified the organization of the output type files in the recently released Outlines v1. `locale` is now a directory containing one file per country.
Yes, I think it's not an issue on Outlines's side. To be sure, could you please provide the version of vLLM and the command used to launch the server?
Great idea! If you could update the content of `llm.txt` that would be awesome too. For the part to add to the documentation for human conception, I think it could...
Adding `reasoning: str` to the top of the type is when you're using a Pydantic class to represent a JSON schema, right?