vllm sampling_params not supported in generate.json

[soumyasmruti]

What behavior of the library made you think about the improvement?

when using generate.json with vllm it doesn't support sampling_params.

Getting the following error

prediction failed because json() got an unexpected keyword argument 'sampling_params'

How would you like it to behave?

Sound be able to do,

  generator = outlines.generate.json(
                        self.model,
                        json_schema,
                        sampling_params=self.sampling_params
                    )