Only in nn.Module: Trying to backward through the graph a second time, but the buffers have already been freed

autograd
1

I have a function that solves a fairly simple optimization problem in the following way:

def my_func(input):
    error = 1
    z = torch.autograd.Variable(some tensor, requires_grad=True)
    opt = torch.optim.SGD([z], lr=0.01)
        while error > 1E-4:
             error = some calculation of input and z 
             with torch.no_grad():
                  opt.zero_grad()
                  error.backward()
                  opt.step()

If I call this function, everything works fine and I don’t get any errors.
However, I need to call this function in forward() of an nn.Module that I wrote. As soon as I call the my_func from within forward(), I receive the following error:

RuntimeError: Trying to backward through the graph a second time, but the buffers have already been freed. Specify retain_graph=True when calling backward the first time.

Note: I do not need to compute gradients w.r.t. the output of my_func.