Reproduce the results

[smallwangzi]

I would greatly appreciate it if you could enlighten me on the difference between running commands with OMP and IMP, respectively. Thank you.

[shivank21]

Thank you for your interest in our work
The main difference between OMP(One Shot Magnitude Pruning) and IMP(Iterative Magnitude Pruning) pruning running commands is that for OMP pruning, you need to prune the model a constant number of times(2). Whereas in IMP pruning, we run it for a larger number of time stamps. You can find the exact details here

[smallwangzi]

Hello, I would like to ask a question regarding Class-wise forgetting. There is a significant difference between the results I obtained and those described in the paper. Can you help me out and tell me what the problem is?

python -u main_forget.py --seed=1 --gpu 0 --data ./datasets/cifar10 --dataset cifar10 --save_dir './_results/cifat10' --mask ./omp/cifar10/1model_SA_best.pth.tar --print_freq=10 --unlearn FT --class_to_replace 0 --unlearn_epochs 10 --unlearn_lr 0.01