Scipy into PyTorch

autograd
1

I am attempting to integrate this scipy function into PyTorch:
https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.UnivariateSpline.html

I’ve been looking at this guide:
https://pytorch.org/tutorials/advanced/numpy_extensions_tutorial.html

This is the code I have written so far:

from scipy.interpolate import UnivariateSpline
import torch
from torch.autograd import Function
class SPLFunction(Function):

@staticmethod
def forward(ctx, input, s):
    np_input = input.detach().numpy()
    s = s.detach()
    
    x, y = np_input[0], np_input[1]

    spl = UnivariateSpline(x, y,s=s)
    result = spl(y)
    ctx.save_for_backward(input, s)

    return torch.as_tensor(result, dtype=input[0].dtype)

@staticmethod
def backward(ctx, grad_output):
    input, s = ctx.saved_tensors
    np_input = input.detach().numpy()
    s = s.detach()
    
    x, y = np_input[0], np_input[1]
    spl = UnivariateSpline(x, y,s=s)
    df = spl.derivative(x)
    df = torch.from_numpy(df*grad_output)
    
    return(df)

However, this does not seem to be working. I am getting a variety of errors…if anyone has experience integrating scipy functions in to PyTorch, any tips in the right direction would be hugely appreciated!