Multi$-$head attention, attention flow, and attention rollout.

$($a$)$ Given queries Q$,$ keys K$,$ and values V$,$ explain how multi$-$head attention computes the

output features.

$($b$)$ Explain how to use multi$-$headed attention described in part $($a$)$ to compute multi$-$head selfattention.

$($c$)$ Describe attention flow and attention rollout