Brief:

Before you read this article, you can skim over the page:

Attention Structure

Self-Attention Structure

Multi-Head Attention Structure

Reference:

自然语言处理中的自注意力机制（Self-attention Mechanism）

Self Attention和Multi-Head Attention的原理和实现_陈建驱的博客-CSDN博客_multihead self-attention

Attention机制详解（二）--Self-Attention与Transformer