yo, which tokenizer to use for K2 text completion?

no matter which one I choose it works