llm-autotune

Tunes LLM serving configurations overnight on your own GPUs: search spaces, pluggable search policies, benchmarked by LLMBench, winners promoted as merge requests

GitHub

Official source

Start locally with the CLI guide