Json mode in offline batch inference

[amritap-ef]

Hi,

Is it possible to have a custom decoding config for outlines / lm-enforcer json format per sample in a batch in offline inference mode?

I have a bunch of samples that I want to enforce json mode for but the json structure is different per sample, and as per the docs it seems that its only possible to pass in a single decoding config to be used for all prompts?