transformer-architecture

Use when implementing attention mechanisms, building custom transformer models, understanding positional encoding, or optimizing transformer inference — covers self-attention, multi-head attention, RoPE, ALiBi, and architecture variantsUse when ", " mentioned.