Qwen 新模型与美团 LongCat 均用 N-gram 架构绕开 HBM 短缺
bookwormengr · x · 2026-09-21
The author argues model architectures are shaped by available hardware: Qwen-3.8-Flash-Next also uses N-gram, a technique Meituan's LongCat lab independently invented. Labs without enough HBM work around it with techniques like N-gram and WideEP to reduce HBM needs — current architectures are not necessarily optimal.
「Infra」频道最新
- 本地频繁换模型不用丢缓存:SGLang hicache 三层 KV 缓存实战 — TheZachMueller · 2026-09-22
- 埃及 AI 生态达生产级规模，4 亿美元数据中心落地 — nordicinst · 2026-09-22
- RTX Pro 6000 还是 3090?本地 LoRA 训练算力经济账对比 — big-in-jap · 2026-09-21
- ComfyUI 租卡实测:Modal 冷启动 35 秒、免费 1TiB 存储胜出 RunPod — ronalder100 · 2026-09-21
- Meta 携手 Arm 开发新一代数据中心 CPU「Arm AGI CPU」 — bookwormengr · 2026-09-21
- Starlink 席卷拉美农村教育：数万座天线连接数十万学生 — XFreeze · 2026-09-21