圖像來源，SCREENSHOT
:first-child]:h-full [&:first-child]:w-full [&:first-child]:mb-0 [&:first-child]:rounded-[inherit] h-full w-full
，推荐阅读91视频获取更多信息
comes with two stacks - the second one being called the break
特点：通过门控机制控制信息流，增强非线性表达。 优点： 适合序列建模、控制性强。 常用于： Transformer FFN、语言模型。