Checklist

• 1. If the issue you raised is not a feature but a question, please raise a discussion at https://github.com/sgl-project/sglang/discussions/new/choose Otherwise, it will be closed.
• 2. Please use English, otherwise it will be closed.

Motivation

as titled

https://github.com/NVIDIA/TensorRT-Model-Optimizer is the de facto LLM quant library for fp8 and fp4, supported in both TensorRT LLM and SGLang. We will consider changing all current fp8, fp4 doc, CI, unit test, etc. to default to ModelOpt's checkpoint

ref https://huggingface.co/nvidia

Related resources

No response