Post
土豆哥
anotherbug.com
did:plc:halfjkmxtkrwqe5ghdedqyvn
周末如果只读一条线
读推理
vLLM 怎么把吞吐做上去
持续批处理怎么把延迟和吞吐一起扛
KV cache 怎么在服务里落地
我自己的排法
先算术
再系统
再部署
目录太长没人看完
链接丢在下面
www.aleksagordic.com/blog/vllm
https://www.anyscale.com/blog/continuous-batching-llm-inference
https://kipp.ly/p/transformer-inference-arithmetic
https://developer.nvidia.com/blog/mastering-llm-techniques-inference-optimization
https://developer.nvidia.com/blog/mastering-llm-techniques-inference-optimization/
2026-09-14T00:47:03.665Z