Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis - View it on GitHub