CODEWITHSUNDEEP
pytorchautograd
Juan De Dios Rojas
Juan De Dios Rojas

Reputation: 25

Vector functions in pytorch to apply autograd to them

If I have a tensor x.

Can I define a vector function, say f(x) = (3x, x+2), and obtain its derivative df/dx?

In a nutshell: I want a way to define such a vector function, from which I can get its gradient.

Upvotes: 0

Views: 856

Answers (1)

Ivan
Ivan

Reputation: 40628

You can do so using torch.autograd.functional.jacobian, providing the function and input:

>>> jacobian(lambda x: (3*x, x+2), inputs=torch.tensor([3.]))

In this case the result is df/dx = (3, 1) for all x.

Upvotes: 1

Related Questions