Self-Attention and Transformer Architecture: BERT, GPT, and Multi-Head Attention