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

Very High Loss #5

Open
vinevix opened this issue Dec 23, 2022 · 1 comment
Open

Very High Loss #5

vinevix opened this issue Dec 23, 2022 · 1 comment

Comments

@vinevix
Copy link

vinevix commented Dec 23, 2022

Hi, I've read your paper and I was trying to train your model. Despite I didn't change hyperparameters or any other model component, I get a very high Loss: that's my result after 15 epochs

epoch 4 average x_t_loss, x_1_loss, prob_loss, val losses: 13.204826354980469, 12.088323593139648, 3156.693359375, 12.881041526794434, 11.268576622009277, 2969.4931640625

Do you know what might be the problem?

@xu-shitong
Copy link
Owner

I suggest you start with the provided Flickr8k dataset clip feature and train for 4-5 epochs, to verify the model is optimized; or use the model to perform inference and see if the model was trained successfully but just with wrong loss calculation.

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