Bicubic interpolation backpropogation

I want to know if the bicubic interpolation can be implemented in pytorch in mathematics.

And I find that there are upsampling functions in pytorch but except for bicubic.
Someone tells me that the bicubic interpolation is not differential but I don’t know the reason. In my knowledge, there is no big difference between bicubic and bilinear.

Finally, any related material is appreciated.