Closed hui-vigil closed 1 year ago
This is a method defined by the super class, torch.autograd.Function
.
Basically, you define the forward and backward pass for the operation and then using .apply()
ensures that PyTorch's autograd keeps track of all the graph info / gradients correctly.
hi,UpFirDn2d.apply( inp, kernel, (up, up), (down, down), (pad[0], pad[1], pad[0], pad[1]) ) but i didn't found the method 'apply' in class UpFirDn2d or Function, could you tell where the 'apply' defined, thank you