# Custom loss function RuntimeError tensor does not have a grad\_fn

**URL:** <https://discuss.pytorch.org/t/custom-loss-function-runtimeerror-tensor-does-not-have-a-grad-fn/104247>\
**Category:** autograd\
**Created:** [November 26, 2020, 7:30pm UTC](https://discuss.pytorch.org/t/custom-loss-function-runtimeerror-tensor-does-not-have-a-grad-fn/104247 "2020-11-26T19:30:02Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![abdelhaqo](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/a/e9bcb4/32.png) [@abdelhaqo](https://discuss.pytorch.org/u/abdelhaqo)\
**Post date:** [November 26, 2020, 7:30pm UTC](https://discuss.pytorch.org/t/custom-loss-function-runtimeerror-tensor-does-not-have-a-grad-fn/104247/1 "2020-11-26T19:30:02Z")

</div>

Trying to utilize a custom loss function and getting error ‘RuntimeError: element 0 of tensors does not require grad and does not have a grad\_fn’. Error occurs during loss.backward()

I’m aware that all computations must be done in tensors with ‘require\_grad = True’. I’m having trouble implementing that as my code requires a nested for loop. Is there a way to create an empty tensor and append it? Below is my code.

```auto
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

```

---

<div class="post-metadata">

**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\
**Post date:** [November 28, 2020, 7:21am UTC](https://discuss.pytorch.org/t/custom-loss-function-runtimeerror-tensor-does-not-have-a-grad-fn/104247/2 "2020-11-28T07:21:24Z")

</div>

You are breaking the computation graph by recreating tensors and the deprecated `Variables`.  
To keep the computation graph intact you would need to apply operations on tensors only, so remove:

```python
... Variable(output),Variable(target)
torch.tensor(error_diff,requires_grad=True)
loss = Variable(loss)

```

etc.