text-generation-inference

fix(server): use server tokenizer as gt

#128

Merged