Analysis of Classifier-Free Guidance Weight Schedulers

TL;DR

This paper investigates different weight scheduling strategies for Classifier-Free Guidance in diffusion models, revealing that simple increasing schedules improve performance and complex ones need task-specific tuning.

Contribution

It provides the first comprehensive analysis of CFG weight schedulers, demonstrating the effectiveness of simple monotonic schedules and the limitations of complex, non-generalizable ones.

Findings

Simple monotonic weight schedulers improve CFG performance.

Complex schedulers can be optimized but lack generalization.

Minimal code changes can implement effective 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.