user
Explain the following
---
assistant
= 1e-4
x = torch.zeros(3, requires_grad=True)
y = (x + t).sum()
y.backward()
print(x.grad)
---
Why does the gradient print exactly ones despite the epsilon perturbation?
Explain the following
---
= 1e-4
x = torch.zeros(3, requires_grad=True)
y = (x + t).sum()
y.backward()
print(x.grad)
---
Why does the gradient print exactly ones despite the epsilon perturbation?