High computational time for UADA and other attack methods

[Jianyi2004]

I am currently running the UADA and the other two attack methods (UPA and TMA) as described in the paper. However, each run is taking approximately 50 hours, which seems excessively long. Could you please advise if this is expected behavior or if there might be an issue with my setup?

Details:

• Hardware: I am using an NVIDIA A100 GPU.
• Configuration: I have followed the setup instructions provided in the paper and the associated code repository.

Questions:

• Is a 50-hour runtime per attack method expected for the UADA, UPA, and TMA?
• Could there be a specific configuration or optimization that I might be missing?
• Are there any known issues or recommendations for reducing the computational time?

Thank you for your assistance.

[William-wAng618]

Thank you for your interest in our work. We have updated the code, and for UADA we now support DDP, which should help optimize the runtime in your case. Additionally, for evaluation, we have uploaded all generated patches for testing (Check in folder adversarial_patches).