Question

我的total_loss为-

一个BCELoss
交叉熵损失
用于图像渐变的自定义损失函数。

我面临的问题是在第一批处理之后，一些权重被更新为nan，导致所有输出均为nan。如果去除梯度损失，则可以正常工作。我发现梯度损失的分母变为0，这引起了问题。为了解决这个问题，我将所有0的分母（梯度量）替换为1。但是这样做只会计算第一个反向支撑的数值损失，而下一个结果将以nan计。

def refineLoss(pred,target,boundary_mask,cuda):

    #sobel_x and sobel_y represent the filters for x and y direction
    sobel_x = torch.tensor([[+1, 0, -1], [+2, 0, -2], [+1, 0, -1]], requires_grad=False,dtype = torch.float)
    sobel_y = torch.tensor([[+1, +2, +1], [0, 0, 0], [-1, -2, -1]], requires_grad=False,dtype = torch.float)
    if cuda:
        sobel_x,sobel_y = sobel_x.cuda(),sobel_y.cuda()
        boundary_mask = boundary_mask.cuda()
    sobel_x = sobel_x.view((1,1,3,3))
    sobel_y = sobel_y.view((1,1,3,3))


    #gradients in the x and y direction for both predictions and the target transparencies
    G_x_pred = F.conv2d(pred,sobel_x,padding = 1)
    G_y_pred = F.conv2d(pred,sobel_y,padding = 1)
    G_x_target = F.conv2d(target,sobel_x,padding = 1)
    G_y_target = F.conv2d(target,sobel_y,padding = 1)

    #magnitudes of the gradients
    M_pred = torch.sqrt(torch.pow(G_x_pred,2)+torch.pow(G_y_pred,2))
    M_target = torch.sqrt(torch.pow(G_x_target,2)+torch.pow(G_y_target,2))

    #see images
    saveimage(M_pred[0],"pred_magnitude")
    saveimage(pred[0],"pred_confidence")
    saveimage(M_target[0],"image_magnitude")


    #taking care of nans
    M_pred = (M_pred==0.).float() + M_pred
    M_target = (M_target==0.).float() + M_target

    # Lcos = (1-v_pred*v_target)*Magnitude_pred
    Lcos = (1-torch.abs((G_x_pred/M_pred)*(G_x_target/M_target)+(G_y_pred/M_pred)*(G_y_target/M_target)))*M_pred

    #Lmag = max(lambda*M_target-M_pred,0)
    lambd = 1.5
    Lmag = lambd*M_target-M_pred
    Lmag[Lmag<0] = 0

    gamma_1 = 0.5
    gamma_2 = 0.5

    #total gradient loss
    total_loss = (gamma_1*Lcos+gamma_2*Lmag)*boundary_mask

    return total_loss.mean()

在第一批反向支撑后总损失为nan

0 个答案: