Skip to content
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

About problems in the training process #57

Open
18222why opened this issue Oct 15, 2024 · 1 comment
Open

About problems in the training process #57

18222why opened this issue Oct 15, 2024 · 1 comment

Comments

@18222why
Copy link

Dear professor, I now have a very strange problem. When I was running the fine-tuning, an unexpected problem occurred. The accuracy rate remained at 78.6 and would not be changed until 100epoch.But it has never happened before, and it will train 10 epochs before it will reach 78.6. Is my model overfitting.
ss

@LTH14
Copy link
Owner

LTH14 commented Oct 15, 2024

It seems your training loss has not decreased. Maybe it is because the VQ tokenization step loses too much information about the image. For a destructive check, you can try to use a very large learning rate (or a very small one) to see whether the accuracy still remains constant. You can also try to evaluate the accuracy before any training.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants