KagamineLenOffical

@zou3519 seems it still won't work if I pass xx_ to the function. Here is the code: ```python import torch from functorch import jacfwd import dogleg class FunctionWrapper(object): def __init__(self,...

What does the mutation here refer to? In my understanding, for the first case, since `jacfwd` can to which args to get the jacobian w.r.t.(not like `functional.jacobian`), the follow code...

Thanks for you reply! Here is a sample version to reproduce this situation: ```python import torch from functorch import jacfwd if __name__ == '__main__': x_replace = torch.tensor([4.,5.,6.]) input_x = torch.tensor([10.,...