均方差损失函数
nn.MSELoss(),当输入两个矩阵时,默认输出标量tensor(torch.Size([])),是两个矩阵对应位置平方差后的平均值
regression_loss = torch.nn.MSELoss(reduction='none') # none 不求平均 # 默认为mean #sum
inputs = torch.tensor([1.,2.])
target = torch.tensor([2.,5.])
loss = regression_loss(inputs, target)
print(inputs)
print(target)
print(loss)
输出
tensor([1., 2.])
tensor([2., 5.])
tensor([1., 9.])
torch.Size([2])