in higher/optim.py [0:0]
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
_warnings.warn(
"Differentiable Rprop (correctly) yields zero second order "
"gradients, as only the sign of the gradient is used in updates. "
"Future versions will offer higher order gradients based on a "
"continuous relaxation of the forward pass."
)