-
Notifications
You must be signed in to change notification settings - Fork 6.3k
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.
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
How do I understand if the training is good? #1096
Comments
Your loss plots look normal. Usually, cycle-consistency loss and identity loss decrease during training, while GAN losses oscillate. To evaluate the quality or detect overfitting/underfitting. you need to apply additional evaluation metrics to your training and test images. The metric is task-specific. See more discussion at #730. |
Hi GZeta95, |
A bit late to the party, in reply to GZeta95's message: I used the last row of each epoch. How I plotted the same (on my own dataset ofcourse):
|
Hi, thanks for your great work. I have trained my datasets with cycle_gan model for 400 epoch. I have generated results with different epochs and qualitatively they have been similar since the 100 epoch, that is, there are samples that are better at the end of the 100 epoch and others at the end of the 400 epoch. So i have plotted the loss functions from loss_log.txt and these are the results:
Which represent the values of functions at the end of each epoch.
Functions oscillate.
What values should I expect to know if the training has gone well? How do I choose how many epochs to tarin the net?
How do I understand if the network is in underfitting or overfitting?
Thank you a lot!
The text was updated successfully, but these errors were encountered: