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

Need the reference hyper-parameters and output of the default scripts. #47

Closed
jeffchy opened this issue Mar 8, 2024 · 1 comment
Closed

Comments

@jeffchy
Copy link

jeffchy commented Mar 8, 2024

I ran the training scripts (using DiT-S/8 by default) successfully, and the loss curve is shown below.
image
However, the sampled results (also using the default sampling parameters with DiT-XL/2 modified to DiT-S/8) are random noises.

Is that because the model is weak?
Could you please provide a recommended setting (hyper-params like the model arch, compression, etc) that we should start with?

@zhengzangw
Copy link
Collaborator

Your number of training steps is quite small. We have included recommended training setting in config, and training loss in our report.

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