What I don’t get in this video is while explaining the building of heads why is everything taken with reference to l’Afrique like each time the W matrices are changed and the question is changed its in reference to l’Afrique and its query is used for multiplication each time to get the right vector, why doesn’t it change? I hope you understand my question
Its just an example to understand the things. We do this same for each word with references to each other words.