For OpenAI Compatible Server mode, Chat API, logits_processors parameter.

[mepwang]

i implemented a logits_processor, which takes tokenizer as init parameter:

class CiteTextLogitsProcessor:
    def __init__(self, target_text, tokenizer):
        self.tokenizer = tokenizer
        self.target_text = target_text

  def __call__(self, input_ids, logits):
    ...

For a llm model in openai compatible server mode, chat api has logits_processors parameters:
https://docs.vllm.ai/en/stable/serving/openai_compatible_server.html#id6

when pass CiteTextLogitsProcessor from openai client, how to construct the constructor objects, especially the tokenizer?