Layer Normalization

Post-LN Transformer

Original transformer architecture where layer normalization is applied after the attention and feed-forward layers, requiring more precise learning rate tuning.

← Tillbaka