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

Evaluation code of the VQVAE #26

Open
whwjdqls opened this issue Oct 17, 2024 · 2 comments
Open

Evaluation code of the VQVAE #26

whwjdqls opened this issue Oct 17, 2024 · 2 comments

Comments

@whwjdqls
Copy link

hi, great work!

I was wondering if there is an evaluation code for the VQVAE.

I tested with another codebase, and I could not get the same results as the paper.

It will be great if you can share the code for evaluating the vqvae.

@exitudio
Copy link
Owner

Hi,
Thank you for your interest. We don't have the evaluation code using test set for VQ. However, you can easily modify the train_vq code.

  1. Change the evaluation from eval to test set in the train_vq.
  2. Load trained model
  3. Move Eval code to before training

P.s. the good result from 1st stage may not transfer 2nd stage. It's better to evaluate the 2nd stage.

@andyl-flwls
Copy link

Hi, Thank you for your interest. We don't have the evaluation code using test set for VQ. However, you can easily modify the train_vq code.

  1. Change the evaluation from eval to test set in the train_vq.
  2. Load trained model
  3. Move Eval code to before training

P.s. the good result from 1st stage may not transfer 2nd stage. It's better to evaluate the 2nd stage.

Hi, I would like ask, i found out from the training log, the FID will get very low early in the training stage, and then, it fluctuate from 0.05 to 0.1 for the remaining training time. Should I stop earlier and shift to stage 2, or continue the training until the end based on your experience? Do you feel even the fid might not be the optimal for stage 1, but longer training time will still make stage 2 performance better?

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

3 participants