Fixed a formula over Attention
This commit is contained in:
parent
65288793ce
commit
e8197d7abc
@ -88,7 +88,7 @@ $$
|
|||||||
\begin{aligned}
|
\begin{aligned}
|
||||||
Head_i = softmax\left(
|
Head_i = softmax\left(
|
||||||
\frac{
|
\frac{
|
||||||
Q_{i} \times K_{i}
|
Q_{i} \times K_{i}^T
|
||||||
}{
|
}{
|
||||||
\sqrt{H}
|
\sqrt{H}
|
||||||
}
|
}
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user