Custom CUDA Or Autograd Functions In Pytorch😃😃

  • Why we have double backward: higher derivatives (https://pytorch.org/tutorials/intermediate/custom_function_double_backward_tutorial.html)