AuraFlow v0.1: a open source alternative to Stable Diffusion 3

AuraFlow v0.1 offers an open-source large scale flow-based generative model for text-to-image creation. It exemplifies the vitality of collaborative AI development and showcases improvements in model scalability, zero-shot learning rates, and re-captioning methodologies. The model is now available for the community to build upon.

• AuraFlow is open-sourced for text-to-image generation.
• It uses MFU, removing many MMDiT layers for efficiency.
• Induces zero-shot learning rates with muP.
• Employs re-captioned datasets for precise instructions.
• Optimizes for larger scale training.