Looking for the Backpropagation in the code

[AmirNoori68]

Hi everyone,
I am trying to locate the backpropagation step in the code but have not been able to find it.
Specifically, I am looking for the computation of the gradients of the loss with respect to those two trainable coefficients in the function:

f(x) = coefficient1 * SiLU(x) + coefficient2 * B_Spline(x).

Could anyone help me identify where this is implemented in the repository?
Thank you!

[FeynmanLogic]

I think you will find it from line 98 from the file MLP.py in the kan folder for the MLP.
KANs don't use traditional backpropagation, they use grid based or block based.