text-generation-inference

6abec14a - feat(server): batch tokenization for flash causal lm (#411)