https://www.hansking.cn/blog A posts created with Next.js and Tailwind.css en-us admin@hansking.cn (HANS) admin@hansking.cn (HANS) Sun, 26 May 2024 10:00:00 GMT https://www.hansking.cn/blog/2026/Mixtral-yu-MoE-xi-shu-zhuan-jia-mo-xing-zen-me-sheng-ji-suan https://www.hansking.cn/blog/2026/Mixtral-yu-MoE-xi-shu-zhuan-jia-mo-xing-zen-me-sheng-ji-suan [{"value":"Mixtral 与 MoE：稀疏专家模型怎么省计算","url":"#mixtral-与-moe稀疏专家模型怎么省计算","depth":1},{"value":"一、密集模型和稀疏模型的区别","url":"#一密集模型和稀疏模型的区别","depth":2},{"value":"二、Mixtral 的核心结构直觉","url":"#二mixtral-的核心结构直觉","depth":2},{"value":"三、参数量为什么容易让人误判","url":"#三参数量为什么容易让人误判","depth":2},{"value":"四、为什么路由器不是一个简单分类器","url":"#四为什么路由器不是一个简单分类器","depth":2},{"value":"五、负载均衡和容量因子","url":"#五负载均衡和容量因子","depth":2},{"value":"六、稀疏计算节省了什么","url":"#六稀疏计算节省了什么","depth":2},{"value":"七、专家并行带来通信挑战","url":"#七专家并行带来通信挑战","depth":2},{"value":"八、Mixtral 让开源生态看见了什么","url":"#八mixtral-让开源生态看见了什么","depth":2},{"value":"九、MoE 的质量不是“专家数量越多越高”","url":"#九moe-的质量不是专家数量越多越高","depth":2},{"value":"十、部署时该怎么选","url":"#十部署时该怎么选","depth":2},{"value":"十一、我的总结：把容量和计算分开，是一次重要转向","url":"#十一我的总结把容量和计算分开是一次重要转向","depth":2}] Sun, 26 May 2024 10:00:00 GMT admin@hansking.cn (HANS) AILLMMixtralMoE稀疏模型