MAML-Pytorch
MAML-Pytorch copied to clipboard
why use custom grad clip function?
https://github.com/dragen1860/MAML-Pytorch/blob/98a00d41724c133bd29619a2fb2cc46dd128a368/meta.py#L41
hey, I'm confused why you didn't use the standard grad clipping function nn.utils.clip_grad_norm(parameters, max_value) instead of implementing clip_grad_by_norm_?