知识卡片
注意力块按计算职责连接
结构图:
flowchart LR
I[词元ID] --> E[嵌入与位置机制]
E --> A[可见范围内的注意力混合]
A --> R[残差与归一化配置]
R --> F[逐位置非线性前馈]
F --> R2[残差与归一化]
R2 --> O[更多层及词表输出]
图给出模块职责;Pre-LN/Post-LN的具体顺序以目标模型公式为准。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 架构与学习目标
原书出处 · 1 处
- · 查看原书依据
Figure 4-2 depicts the original Transformer architecture, developed in 2017 by Vaswani et al. As we can see in the figure, a Transformer model is typically composed of Transformer blocks stacked on top of each other, called layers. The key components of each block are:
参考来源
- 位置:《Designing Large Language Model Applications》第04章《架构与学习目标》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch04.html)
- 结论依据:摘录中关于注意力块按计算职责连接的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Figure 4-2 depicts the original Transformer architecture, developed in 2017 by Vaswani et al. As we can see in the figure, a Transformer model is typically composed of Transformer blocks stacked on top of each other, called layers. The key components of each block are: