Gradient of X is NoneType in second iteration

autograd
1

Hi, I’m trying to make images which will fool model, but I have some problem with this code. In the second iteration I get TypeError: unsupported operand type(s) for -: 'Tensor' and 'NoneType'
Why is the grad NoneType even if it’s working for the first time?

X_fooling = X.clone()
X_fooling.requires_grad_()
loss_f = torch.nn.MSELoss()

for i in range(1000):
      score = model(X_fooling)
      y = torch.zeros(1000)
      y[target_y] = 1
      loss = loss_f(score, y)
      print(loss)
      loss.backward()
        
      if target_y == torch.argmax(score):
        break

      X_fooling = X_fooling - X_fooling.grad
2

You are replacing the leaf variable:

X_fooling = X.clone()
X_fooling.requires_grad_()

with a non-leaf:

X_fooling = X_fooling - X_fooling.grad

I’m unsure if you want to recreate the leaf variable, but you could use:

X_fooling.detach().requires_grad_()

to recreate the leaf variable.

3

Thank you, I solved the problem with usage of optimizer. I haven’t thought about replacing of the tensor