search

OPTML-Group / Unlearn-Saliency

[ICLR24 (Spotlight)] "SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation" by Chongyu Fan*, Jiancheng Liu*, Yihua Zhang, Eric Wong, Dennis Wei, Sijia Liu
https://www.optml-group.com/posts/salun_iclr24
MIT License
90 stars 12 forks source link

About applying mask in random_label.py #15

Closed TianyunYoung closed 3 months ago

TianyunYoung commented 3 months ago

I find something strange in random_label.py. As seen on line 133, this line is for select those parameters that have gradient and in the parameters that are optimized. However, for the condition n in parameters, n is a string, and parameters is a list of weights, rather than a list of names. So n in paramters will always be False, which makes the mask not applied. Is that true? Or I miss something.

8991715569483_ pic

a-F1 commented 3 months ago

Thank you for your attention and support towards our work, as well as for bringing this issue to our notice. We sincerely apologize for any misunderstanding caused by the oversight during the code cleanup process. Rest assured, we have promptly rectified the bug.

If you have any further inquiries or require additional assistance, please feel free to reach out to us.

TianyunYoung commented 3 months ago

I got it. Thanks for your prompt clarification.