Analysis of Classifier-Free Guidance Weight Schedulers

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Xi, Dufour, Nicolas, Andreou, Nefeli, Cani, Marie-Paule, Abrevaya, Victoria Fernandez, Picard, David, Kalogeiton, Vicky
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909463512875008
author Wang, Xi
Dufour, Nicolas
Andreou, Nefeli
Cani, Marie-Paule
Abrevaya, Victoria Fernandez
Picard, David
Kalogeiton, Vicky
author_facet Wang, Xi
Dufour, Nicolas
Andreou, Nefeli
Cani, Marie-Paule
Abrevaya, Victoria Fernandez
Picard, David
Kalogeiton, Vicky
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2404_13040
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Analysis of Classifier-Free Guidance Weight Schedulers
Wang, Xi
Dufour, Nicolas
Andreou, Nefeli
Cani, Marie-Paule
Abrevaya, Victoria Fernandez
Picard, David
Kalogeiton, Vicky
Computer Vision and Pattern Recognition
Machine Learning
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.
title Analysis of Classifier-Free Guidance Weight Schedulers
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2404.13040