Visual Generation Without Guidance

TL;DR

This paper introduces Guidance-Free Training (GFT), a novel approach for visual generative models that eliminates the need for guided sampling, reducing computational costs while maintaining performance.

Contribution

GFT enables training guidance-free visual models from scratch, matching CFG performance with simpler implementation and lower computational costs.

Findings

GFT achieves comparable or better FID scores across multiple models.

GFT reduces sampling to a single model, halving computational costs.

GFT maintains diversity-fidelity trade-offs similar to CFG.

Abstract

Classifier-Free Guidance (CFG) has been a default technique in various visual generative models, yet it requires inference from both conditional and unconditional models during sampling. We propose to build visual models that are free from guided sampling. The resulting algorithm, Guidance-Free Training (GFT), matches the performance of CFG while reducing sampling to a single model, halving the computational cost. Unlike previous distillation-based approaches that rely on pretrained CFG networks, GFT enables training directly from scratch. GFT is simple to implement. It retains the same maximum likelihood objective as CFG and differs mainly in the parameterization of conditional models. Implementing GFT requires only minimal modifications to existing codebases, as most design choices and hyperparameters are directly inherited from CFG. Our extensive experiments across five distinct visual models demonstrate the effectiveness and versatility of GFT. Across domains of diffusion, autoregressive, and masked-prediction modeling, GFT consistently achieves comparable or even lower FID scores, with similar diversity-fidelity trade-offs compared with CFG baselines, all while being guidance-free. Code will be available at https://github.com/thu-ml/GFT.