rtx-8000

### Your current environment ```text The output of `python collect_env.py` ``` ### How would you like to use vllm I've noticed that using LoRA with rank=256 significantly slows down inference...