用140万参数模型在40万步内跑出ImageNet扩散模型顶尖效果。
Speedrunning ImageNet Diffusion
- 整合分词路由、架构优化和训练策略提升效率
- FID达3.49，KDD为0.319，性能媲美更大模型
- 适合追求高效训练的扩散模型研究者
近期进展显著提升了扩散Transformer的训练效率，但这些方法多独立研究，缺乏系统性结合。本文提出SR-DiT框架，系统集成分词路由、架构改进与训练优化，并基于表征对齐构建基础。仅用140M参数模型，在40万次迭代下无需分类器自由引导即可实现ImageNet-256上FID 3.49与KDD 0.319，性能媲美685M参数模型长期训练结果。通过大量消融实验，识别出最有效的技术组合，揭示协同与冲突关系。研究开源该框架，作为未来研究的可计算基线。
原文摘要 · Abstract (English)
Recent advances have significantly improved the training efficiency of diffusion transformers. However, these techniques have largely been studied in isolation, leaving unexplored the potential synergies from combining multiple approaches. We present SR-DiT (Speedrun Diffusion Transformer), a framework that systematically integrates token routing, architectural improvements, and training modifications on top of representation alignment. Our approach achieves FID 3.49 and KDD 0.319 on ImageNet-256 using only a 140M parameter model at 400K iterations without classifier-free guidance - comparable to results from 685M parameter models trained significantly longer. To our knowledge, this is a state-of the-art result at this model size. Through extensive ablation studies, we identify which technique combinations are most effective and document both synergies and incompatibilities. We release our framework as a computationally accessible baseline for future research.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。