search

ZJUJeffLai / SAW_SSL

MIT License
13 stars 4 forks source link

Inconsistent result on baseline Fixmatch #3

Closed butcher1226 closed 1 year ago

butcher1226 commented 1 year ago

Sorry to bother you, but when reproducing your code, I can't get the same results mentioned in your paper. Even when removing the adaptive weighting, just run Fixmatch, I got the mean acc 55.06, which is much worse than 60.1 in your paper, could you explain why? Thank you.

butcher1226 commented 1 year ago

The setting is CIFAR100, imb_ratio_l 10, imb_ratiol_u 10, num_max 150

ZJUJeffLai commented 1 year ago

Hello, thanks for your message. Our codes are based on this repo: https://github.com/bbuing9/DARP Could you run this repo and see if you can get 60.1? Meanwhile, I'll run my Repo on my side to see the issues. I may not upload the right version of the scripts. Sorry about that.