Using already calculated values in `autograd.functional.jacobian`

[pascal_notsawo (Tikquuss)]

I already made a post about the jacobian computation with pytorch (as well as the gradient, divergence and laplacian).
Have a look at it, it might be useful