模拟源文件：Attention Is All You Need (Vaswani et al., 2017)
知识结构思维导图
graph TD
A[Transformer 架构] --> B(核心机制)
A --> C(模型结构)
A --> D(训练技巧)
B --> B1[Self-Attention]
B --> B2[Multi-Head Attention]
B --> B3[Scaled Dot-Product]
C --> C1[Encoder-Decoder]
C --> C2[Positional Encoding]
C --> C3[Feed-Forward Networks]
D --> D1[Residual Connections]
D --> D2[Layer Normalization]
D --> D3[复杂度优势]
style A fill:#f3f4f6,stroke:#333,stroke-width:2px
style B fill:#d1fae5,stroke:#059669
style C fill:#dbeafe,stroke:#2563eb
style D fill:#ede9fe,stroke:#7c3aed