a$)$ Consider a text sequence with $3$ words: such that each word is represented by a $2-$d vector as given in matrix $x.$ Show

by step calculations for computing the scaled dot$-$product based self attention scores using the Wq$.$ Wk and Wv matrices given below

Marks$]$

b$)$ i$)$ Mention and justify the different types of attention mechanisms used in Transformer architecture.

$[2$ Marks$]$

ii$)$ Justify the selection of Bahdanau attention used and the disadvantages due to this selection. Is there any alternative to this attention $?$

$[2$ Marks$]$