Multi-Head Attention

Contextual Embedding

Enriched vector representation generated by the attention mechanism that incorporates contextual information from the entire sequence for each element.

← 뒤로