L1Loss MSELoss 都没问题,但是 HingeEmbeddingLoss 却总报错 说不能求梯度
#criterion = nn.MSELoss()
#criterion = nn.L1Loss()
criterion = nn.HingeEmbeddingLoss()
发现 Hinge损失函数还是区分了
-
Input: (*)(∗) where *∗ means, any number of dimensions. The sum operation operates over all the elements.
-
Target: (*)(∗) , same shape as the input
-
Output: scalar. If
reduction
is'none'
, then same shape as the input
保证模型输出的变量在第一个就可以了,第二个是Label/Target
loss =criterion(out, data.y)