text-generation-inference

fbeb1c44 - fix: follow base model for tokenizer in router (#1424)