I can not recover the image from the gradients.

mit-han-lab / dlg

[NeurIPS 2019] Deep Leakage From Gradients

https://dlg.mit.edu/

MIT License

396 stars 105 forks source link

I can not recover the image from the gradients. #9

Closed 6lyc closed 1 year ago

6lyc commented 1 year ago

Why can't I reproduce the effect of the paper? According to your code, the image I get is just some disordered pixel blocks, even if I increase the number of iterations.

6lyc commented 1 year ago

It should set a little learning rate, otherwise the dlg model will not converge.

FryLcm commented 9 months ago

应该设置一点学习率，否则dlg模型不会收敛。

能详细说下怎么解决吗

6lyc commented 9 months ago

See https://github.com/6lyc/DLG-and-its-defense.git for more details. @FryLcm

FryLcm commented 8 months ago

有关更多详细信息，请参阅https://github.com/6lyc/DLG-and-its-defense.git 。@FryLcm

感谢您的回复！