[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"topic-zh-pagedweight-moe-":3},{"cluster":4,"timeline":16},{"id":5,"slug":6,"title":7,"pinned":8,"status":9,"summary":10,"category":11,"language":12,"created_at":13,"merged_into":14,"article_count":15,"first_seen_at":13,"last_updated_at":13},"2233b031-a1d6-4974-8455-2dcb22777c1a","pagedweight-moe-","PagedWeight 動態量化 MoE 省顯存",false,"active","PagedWeight 在推理時動態量化 MoE 權重，換出更多 GPU 記憶體給 KV cache，且維持接近 FP16 的品質。","research","zh","2026-07-20T06:02:28.232877+00:00",null,1,[17],{"id":18,"slug":19,"title":7,"summary":10,"category":11,"image_url":20,"cover_image":20,"published_at":21,"is_canonical_seed":22},"2b9e6590-18ac-46f7-ae85-4a2eecabe0b4","pagedweight-moe-serving-dynamic-quantization-zh","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784527374591-fxj3.png","2026-07-20T06:02:26.998+00:00",true]