text-generation-inference

4f460e5b - feat(server): improve max tokens calculation