noisify_pairflip missing

[xiaoyihit]

noisify_pairflip missing
noisy_long_tail_CIFAR.py, line 142

corruption_list = {
        'uniform': uniform_corruption,
        'flip1': noisify_pairflip,  # missing
        'flip2': flip2_corruption,
    }

Should I change 'noisify_pairflip' into 'flip1_corruption'?

Another issue is that I fail to reproduce the results on CIFAR10 and ended up with final acc 86.42±0.27% which is lower than 88.45±0.15% with command:
python robot_main.py --dataset cifar10 --corruption_ratio 0.5 --analyze --meta_lr 5e-2 --start_correction 0 --runs_name 0.5_forward_rce_uniform --loss forward --print_predictions --start_updating_T 20 --max_epoch 85 --outer_obj rce --corruption_type uniform
robot_main_20230417_1602_ori_uniform.log
robot_main_20230417_2206_ori_uniform.log
robot_main_20230418_0016_ori_uniform.log
robot_main_20230418_1140_ori_uniform.log
robot_main_20230418_1141_ori_uniform.log

91.86±0.21% with
python robot_main.py --dataset cifar10 --corruption_ratio 0.45 --analyze --meta_lr 5e-2 --start_correction 0 --runs_name 0.45_forward_rce_flip --loss forward --print_predictions --start_updating_T 20 --max_epoch 85 --outer_obj rce --corruption_type flip1
robot_main_20230417_1600_ori_flip1.log
robot_main_20230417_2205_ori_flip1.log
robot_main_20230418_0017_ori_flip1.log
robot_main_20230418_1705_ori_flip1.log
robot_main_20230418_1707_ori_flip1.log

Although this result still surpasses VolMinNet by a large margin, I face trouble reproducing results given in the paper.

Plz help, thx in advance.