Abra: Scaling Diffusion Image Training
Scaling laws for text-to-image diffusion models reveal predictable compute-optimal training requiring far more data per parameter than language models, with robust overtraining beh…
Hugging Face · Daily Papers
·Kyle Chickering, Wei-An Lin
·
·▲ 7 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Kyle Chickering, Wei-An Lin, Swayam Bhanded, Dan Saunders, Akshat Tripathi, Jiaming Song
- 7 upvotes da comunidade
- Temas: scaling laws, text-to-image diffusion models, flow-matching transformers, compute-optimal, Chinchilla, CFG
Resumo
Resumo original (em inglês), extraído do paper:
Scaling laws for text-to-image diffusion models reveal predictable compute-optimal training requiring far more data per parameter than language models, with robust overtraining behavior and universal curve shapes.