vLLM新版本把Model Runner V2设成了默认，推理更快更稳，还新增了流解析和推测解码，想优化部署性能的话可以关注。
vLLM v0.25.0 正式发布，包含 558 次提交和 232 位贡献者。Model Runner V2 现已成为所有密集模型的默认执行路径，同时旧版 PagedAttention 实现被移除。Transformers 后端的运行速度已与原生 vLLM 持平。新增统一流解析引擎，以及支持跨异构词汇表的通用推测解码（TLI），并引入 DSpark 和 DFlash 起草器。新模型支持包括 Hy3 和 Unlimited OCR。
vLLM v0.25.0 is out! 558 commits from 232 contribu…
vLLM v0.25.0 is out! 558 commits from 232 contributors (64 new). 🎉
Highlights: Model Runner V2 is now the default for all dense models, the legacy PagedAttention implementation is retired, the Transformers backend now runs as fast as native vLLM, a new unified Streaming Parser Engine, universal speculative decoding across heterogeneous vocabularies (TLI) plus new DSpark and DFlash drafters, and new models including Hy3 and Unlimited OCR.
Thread 👇