← 返回卡片

原书出处

Designing Large Language Model Applications

First Edition · 核验版本的摘录快照 · 导入于 2026-11-14T13:54:13.935Z

A key component of the MoE architecture is the gating function. The gating function helps decide which expert is more suited to process a given input. The gating function is implemented as a weight applied to each expert. The experts are typically added to the feedforward component of the Transformer. Therefore, if there are eight experts, then there will be eight feedforward networks instead of

此预览只包含定位所需的摘录和上下文。

原书定位详情

来源:a19c7933-4eda-4176-aebd-30fe7e1ad292

正文位置:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch04.html

EPUB SHA-256:1b4d2347fec2f949fcddfbaf4fad597a1ae20887a096cfae64c59a619b4781bf