Transformer

![[Pasted image 20240607182037.png]]

bathcNorm是每个特征归一化，layerNorm是每个样本归一化 ![[Pasted image 20240607232037.png]]

学习资源

https://www.bilibili.com/video/BV1pu411o7BE/?spm_id_from=333.337.search-card.all.click&vd_source=7ef7dff4b509c161e6b86a796dbad2c5

博客 The Illustrated Transformer