Add OpenAI to LogitsGenerator

[ulfaslak]

This is tiny, and I might be wrong in case OpenAI (since it's basically an API wrapper) is in fact NOT a LogitsGenerator.

But adding it seems logical, and I want it to be in there for my use case (which is that I have a function that creates a model, and with this change I can specify the return type to be LogitsGenerator).

• We should be able to understand what the PR does from its title only;
• There is a high-level description of the changes;
• If I add a new feature, there is an [issue][issues] discussing it already;
• There are links to all the relevant issues, discussions and PRs;
• The branch is rebased on the latest main commit;
• Commit messages follow these [guidelines][git-guidelines];
• One commit per logical change;
• The code respects the current naming conventions;
• Docstrings follow the [numpy style guide][docstring-guidelines];
• pre-commit is installed and configured on your machine, and you ran it before opening the PR;
• There are tests covering the changes;
• The documentation is up-to-date;

[lapp0]

Could you help me understand what you're using the LogitsGenerator for in your codebase? Just generally to signify that it's an OutlinesModel?

Perhaps we should change this type to OutlinesModel?

[ulfaslak]

Could you help me understand what you're using the LogitsGenerator for in your codebase? Just generally to signify that it's an OutlinesModel?

Perhaps we should change this type to OutlinesModel?

Yeah, so I have a function that accepts an argument model of type outlines.models.LogitsGenerator | outlines.models.OpenAI. I am using beartype for runtime type checking, so the outlines.models.LogitsGenerator is a convenient type because it allows all the model types that might get passed to the function.

I searched the codebase, can't find an OutlinesModel class anywhere. Maybe I misunderstand you?

[lapp0]

Thanks for helping me understand.

We don't have an OutlinesModel, but that'll likely be its name after a refactor.

[ulfaslak]

Roger. Yes something like that will be useful. Also for cases outside of runtime type validation, I can think of many cases where you'd want to do one thing if the model is a ready-to-use OutlinesModel, and another thing if it's not (e.g. a string: then create the model)