SGLang 整合 Hugging Face Transformers 後端：大幅提升模型相容性與開發彈性

SGLang (Structured Generation Language) is a high-performance LLM inference and serving framework developed by the LMSYS team, renowned for its efficient execution and structured output capabilities. In the past, SGLang relied primarily on its heavily optimized native execution engine (SRT, similar to vLLM) to pursue maximum throughput. However, this also meant that whenever the community released a new model architecture, developers had to wait for the SGLang team to write dedicated CUDA kernels and optimization code before they could use it.