Tian0426 / CL-HAR

56 stars 10 forks source link

Any ideas about the inconsistency #8

Open Berken-demirel opened 1 year ago

Berken-demirel commented 1 year ago

Hello there,

Great framework! Thanks a lot for sharing!

I have a question regarding the results of the repository you published on GitHub. I am running the code with the following command:

`python main.py --framework 'simclr' --backbone 'DCL' --dataset 'ucihar' --aug1 'rotation' --aug2 'perm' --n_epoch 120 --batch_size 256 --lr 3e-3 --lr_cls 0.03 --cuda 3 --cases subject --target_domain 3
`

However, I am getting different results compared to the published results (with the below output ). I was wondering if you could help me figure out what the issue might be. I appreciate your time and effort. Thank you.

Epoch : 119 Saving model at 119 epoch to results/lincls_try_scheduler_simclr_pretrain_ucihar_eps120_lr0.003_bs256_aug1rotation_aug2perm_dim-pdim128-128_EMA0.996_criterion_NTXent_lambda1_1.0_lambda2_1.0_tempunit_tsfm119.pt epoch train loss : 4.5841, train acc : 63.7500 epoch test loss : 2.0348, test acc : 58.0442, miF : 58.0442, maF : 52.7896 tensor([[59., 1., 0., 0., 0., 0.], [32., 17., 3., 0., 0., 0.], [27., 14., 4., 0., 0., 0.], [ 1., 0., 0., 12., 28., 9.], [ 2., 0., 0., 9., 45., 0.], [ 0., 0., 0., 3., 4., 47.]]) tensor([0.9833, 0.3269, 0.0889, 0.2400, 0.8036, 0.8704])