New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

What's the pytorch version #5

Open

yihao31 opened this issue Mar 11, 2019 · 3 comments

yihao31 commented Mar 11, 2019

I trained the model with pytorch 0.4.0 in Python3.6.3. The loss_G and loss_identity show a very large number.

Owner

sshkhr commented Mar 16, 2019

I also used pytorch 0.4.0
I will check this issue after the ICCV deadline and get back to you

yuzi56 commented Sep 26, 2019

I got the same problem. The loss has been rising and finally turned into NaN.
Could you help me address it? Thanks a lot.
Epoch --------------- 1
Epoch 001/2000 [0001/5071] -- loss_G: 15.6972 | loss_D: 1.3689 | loss_triplet_rgb: 2.6829 | triplet_loss_ir: 1.2321 | loss_identity: 11.7822 | loss_total: -15.6287
-Epoch 001/2000 [0002/5071] -- loss_G: 16.1317 | loss_D: 1.3754 | loss_triplet_rgb: 2.3103 | triplet_loss_ir: 1.9274 | loss_identity: 11.8940 | loss_total: -16.0629
-Epoch 001/2000 [0003/5071] -- loss_G: 17.5205 | loss_D: 1.3803 | loss_triplet_rgb: 2.1013 | triplet_loss_ir: 3.4595 | loss_identity: 11.9597 | loss_total: -17.4515
-Epoch 001/2000 [0004/5071] -- loss_G: 74.9674 | loss_D: 1.5213 | loss_triplet_rgb: 22.0515 | triplet_loss_ir: 38.1460 | loss_identity: 14.7699 | loss_total: -74.8913
Epoch 001/2000 [0005/5071] -- loss_G: nan | loss_D: 4976841028734326669312.0000 | loss_triplet_rgb: nan | triplet_loss_ir: nan | loss_identity: 697867652548091466219
Epoch 001/2000 [0006/5071] -- loss_G: nan | loss_D: nan | loss_triplet_rgb: nan | triplet_loss_ir: nan | loss_identity: nan | loss_total: nan

MykolaHavrylov commented Nov 8, 2021 •

edited

Loading

@sshkhr Yeah, I have the same problem

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment