This is official implication for our paper: CroSel: Cross Selection of Confident Pseudo Labels for Partial-Label Learning (Accepted by CVPR 2024)
#train_command for cifar10 with q=0.5
python mix_main.py --dataset cifar10 --arch wideresnet --batch-size 64 --lr 0.1 \
--seed 5 --out cifar10@q05 --partial_rate 0.5 --epochs 200 --gpu_id 7 --sharpen_T 0.5 --use_mix --lambda_cr 4
#train_command for cifar100 with q=0.1
python mix_main.py --dataset cifar100 --arch wideresnet --batch-size 64 --lr 0.1 \
--seed 5 --out cifar100@q01 --partial_rate 0.10 --epochs 200 --gpu_id 6 --sharpen_T 0.5 --use_mix