WindVChen
Hi @tanlingp , In theory, we can of course generate adversarial images in batches. However, since this method currently requires backpropagation through multiple diffusion steps, many intermediate calculation graphs are...
Hi @yujinhan98, Thank you for sharing the article. I appreciate your insight in addressing the memory issue with Segment-wise Forwarding-Backwarding. However, it seems that this approach might prolong the optimization...
Hi @tanlingp , Based on the above information, I am currently unable to conclude what is causing this error. Can you take a look at your `diffusers` version? It might...
Hi @nil0330 , You can take a look at [requirements.txt](https://github.com/WindVChen/DiffAttack/blob/main/requirements.txt) for the versions of some packages. Hope this helps.
Hi @tanlingp , This seems an Internet Connection problem. Try to recheck your connection or manually download the weight from [here](https://huggingface.co/stabilityai/stable-diffusion-2-base/tree/main).
Hi @nil0330 , An alternative way is to download the pre-trained weight manually. I would recommend referring to the answer [here](https://stackoverflow.com/questions/71891127/is-there-any-way-i-can-use-the-downloaded-pre-trained-models-for-timm). Hope this helps.
Hi @tanlingp, Are you inquiring about how to evaluate anti-defense performance? Defense methods have multiple categories, like adversarially trained models, defensive purification, etc. For adversarial-trained models, you simply need to...
To run the code for the defenses listed in the "Robustness on defensive approaches" section, I would recommend checking the respective source repositories. The README files in these repositories usually...
Hi @nil0330 , Our method is optimization-based and doesn't require training using `train.py`. Please refer to [here](https://github.com/WindVChen/DiffAttack#crafting-adversarial-examples) for optimization commands and parameters. You can also find more details in [diff_latent_attack.py](https://github.com/WindVChen/DiffAttack/blob/main/diff_latent_attack.py).
Hi @tanlingp , It does take time. Slow running speed is probably a common problem of most diffusion models at present. For a quick look at the results, maybe you...