Sensitive and Biased scenarios implement question

[win7V]

Dear Author，Thank you very much for sharing the code associated with your paper. I sincerely appreciate your contributions to this field. I am currently learning from your project and have successfully reproduced the backdoor experiment. However, I am encountering difficulties in implementing the experiment under Sensitive and Biased scenarios. I would greatly appreciate it if you could provide some guidance on executing the relevant instructions.
Additionally, I noticed what appears to be an issue with the -perturbed_part selection in your configs/train_config.py file. Should this parameter be updated to include the following
choices= ["mouth", "eye", "nose", "face", "face_except_mouth"] ?
Your suggestions regarding this matter would be highly valued.
Thank you once again for sharing your paper's code!

[Pita-Ten]

Hi, could you please advise on how to reproduce the backdoor experiment? I entered some issues: 1) For example, the unlearn results for the MNIST dataset cannot be achieved; 2) when implementing the CIFAR10 dataset in the baseline, the "unlearn test acc" consistently remains at only 79%. This still shows a deviation from the results in the paper. How should the hyperparameters be adjusted? Is the code complete?