Analysis of Classifier-Free Guidance Weight Schedulers
Abstract: Classifier-Free Guidance (CFG) enhances the quality and condition adherence of text-to-image diffusion models. It operates by combining the conditional and unconditional predictions using a fixed weight. However, recent works vary the weights throughout the diffusion process, reporting superior results but without providing any rationale or analysis. By conducting comprehensive experiments, this paper provides insights into CFG weight schedulers. Our findings suggest that simple, monotonically increasing weight schedulers consistently lead to improved performances, requiring merely a single line of code. In addition, more complex parametrized schedulers can be optimized for further improvement, but do not generalize across different models and tasks.
Submission Length: Regular submission (no more than 12 pages of main content)
Previous TMLR Submission Url: https://openreview.net/forum?id=bGrSfr1kxZ&nesting=2&sort=date-desc
Changes Since Last Submission: Dear Editor,
We have revised our manuscript, originally submitted as No. 3007, and are resubmitting it for your consideration. We have rectified the template problem of the manuscript.
Best regards,
Assigned Action Editor: ~Hanwang_Zhang3
Submission Number: 3074
Loading