[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-推理壓縮":3},{"tag":4,"articles":9},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":8},"9dc918dc-2498-4e0d-a186-b5393a4ec891","推理壓縮",0,null,[10],{"id":11,"slug":12,"title":13,"summary":14,"category":15,"image_url":16,"cover_image":16,"language":17,"created_at":18},"de80c7d2-c4bf-4168-855f-1d5ac9e34151","turboquant-vllm-kv-cache-3bit-storage-zh","TurboQuant 把 vLLM KV 快取壓成 3 位元","我拆 TurboQuant 怎麼把 vLLM 的 KV cache 壓到 3-bit，順手整理成可直接套用的部署判斷與模板。","tools","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779286504704-sisl.png","zh","2026-05-20T14:14:36.149676+00:00"]