Custom loss function error: tensor does not have a grad_fn

Question

尝试使用自定义损失函数并出现错误“RuntimeError: element 0 of tensors does not require grad and does not have a grad_fn”。 loss.backward()

期间发生错误

我知道所有计算都必须在“require_grad = True”的张量中完成。我在实现它时遇到了麻烦，因为我的代码需要一个嵌套的 for 循环。我相信这可能是 for 循环。有没有办法创建一个空张量并附加它？下面是我的代码。

def Gaussian_Kernal(x, mu, sigma):
  p = (1./(math.sqrt(2. * math.pi * (sigma**2)))) * torch.exp((-1.) * (((Variable(x)**2) - mu)/(2. * (sigma**2))))
  return p

class MEE(torch.nn.Module):
  def __init__(self):
    super(MEE,self).__init__()

  def forward(self,output, target, mu, variance):

    error = torch.subtract(Variable(output),Variable(target))
  
    error_diff = []
    for i in range(0, error.size(0)):
      for j in range(0, error.size(0)):
        error_diff.append(error[i] - error[j])

    error_diff = torch.cat(error_diff)
    torch.tensor(error_diff,requires_grad=True)

    loss = (1./(target.size(0)**2)) * torch.sum(Gaussian_Kernal(Variable(error_diff), mu, variance*(2**0.5)))

    loss = Variable(loss)

    return loss

Answer 1

只要你在张量上操作并应用 PyTorch 函数和基本运算符，它应该可以工作。因此不需要用 torch.tensor 或 Variable 包装你的变量。后者已被弃用（我相信从 v0.4 开始）。

The Variable API has been deprecated: Variables are no longer necessary to use autograd with tensors. Autograd automatically supports Tensors with requires_grad set to True. PyTorch docs

我假设 output 和 target 是张量，而 mu 和 variance 是实数而不是张量？那么，output 和 target 的第一个维度就是批次。

def Gaussian_Kernel(x, mu, sigma):
  p = (1./(math.sqrt(2. * math.pi * (sigma**2)))) * torch.exp((-1.) * (((x**2) - mu)/(2. * (sigma**2))))
  return p

class MEE(torch.nn.Module):
  def __init__(self):
    super(MEE, self).__init__()

  def forward(self, output, target, mu, variance):
    error = output - target

    error_diff = []
    for i in range(0, error.size(0)):
      for j in range(0, error.size(0)):
        error_diff.append(error[i] - error[j]) # Assuming that's the desired operation

    error_diff = torch.cat(error_diff)
    kernel = Gaussian_Kernel(error_diff, mu, variance*(2**0.5))
    loss = (1./(target.size(0)**2))*torch.sum(kernel)
    
    return loss

Custom loss function error: tensor does not have a grad_fn

Custom loss function error: tensor does not have a grad_fn

python

machine-learning

deep-learning

recurrent-neural-network

pytorch