arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.15420cs.CVcs.AIcs.LG
收藏

无需引导的视觉生成

Visual Generation Without Guidance

Department of Computer Science \& Technology, Tsinghua University

Huayu Chen, Kai Jiang, Kaiwen Zheng, Jianfei Chen, Hang Su, Jun Zhu

AI总结:

针对视觉生成模型中分类器无引导(CFG)需双模型推理的问题,提出无引导训练(GFT),无需预训练CFG网络可直接从头训练,保留最大似然目标且实现简单,采样仅需单模型、计算成本减半,性能匹配CFG且FID分数相当或更低。

摘要:

分类器无引导(CFG)是各种视觉生成模型中的默认技术,但它在采样期间需要同时从条件模型和无条件模型进行推理。我们提出构建无需引导采样的视觉模型。由此产生的算法——无引导训练(GFT)——在将采样简化为单一模型、将计算成本减半的同时,匹配了CFG的性能。与依赖预训练CFG网络的先前基于蒸馏的方法不同,GFT支持直接从头开始训练。GFT实现简单,它保留了与CFG相同的最大似然目标,主要区别在于条件模型的参数化。由于大多数设计选择和超参数直接继承自CFG,实现GFT只需对现有代码库进行最小修改。我们在五个不同视觉模型上的广泛实验证明了GFT的有效性和通用性。在扩散、自回归和掩码预测建模领域,GFT始终实现了相当甚至更低的FID分数,与CFG基线相比具有相似的多样性-保真度权衡,且完全无需引导。代码将在https://github.com/thu-ml/GFT提供。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑