add third chapter
This commit is contained in:
parent
5498cf5935
commit
f15fde036b
|
|
@ -1000,11 +1000,11 @@ tensor([[2., 2., 0., 2., 2., 0.],
|
|||
>
|
||||
> $$ \text{z_{i}}^{\prime}=\frac{z_{i}}{1-p} \quad \text { (对于未被置零的权重) } $$
|
||||
>
|
||||
>
|
||||
>
|
||||
> $$ \text{softmax}\left(z_{i}\right)=\frac{e^{z_{i}}}{\sum_{j} e^{z_{j}}} $$
|
||||
>
|
||||
>
|
||||
>
|
||||
>
|
||||
> 111
|
||||
|
||||
现在,让我们将 dropout 应用于注意力权重矩阵本身:
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue