This AI Paper from Shanghai AI Laboratory Introduces Lumina-mGPT: A High-Resolution Text-to-Image Generation Model with Multimodal Generative Pretraining