Transformer：一次注意力怎么算？Transformer: Calculate One Attention Step

区分 query、key、value，计算三个权重的和，并解释因果遮罩为什么不能偷看未来。Distinguish queries, keys and values, check that weights sum to one and explain why causal masking hides the future.

先在纸上猜结果，再按按钮检验。Predict on paper before checking with the buttons.

数值是人为设置的单头示例，没有学到词义。完整 Transformer 还包含学习到的投影、多头、位置表示、前馈层、残差等；这里不训练完整模型。These are hand-set numbers for one attention head without learned word meanings. A complete Transformer also has learned projections, multiple heads, position information, feed-forward layers and residuals.

回到课程Back to the lesson