首个基于昇腾MindSpeed MM训练的多模态生成理解统一模型Lumina-DiMOO正式发布

Core Viewpoint - The Shanghai Artificial Intelligence Laboratory, in collaboration with Huawei Ascend, launched the Lumina-DiMOO model at the 2025 World Artificial Intelligence Conference (WAIC), marking a significant advancement in multimodal AI capabilities [1] Group 1: Product Launch - The Lumina-DiMOO model is a next-generation unified model for multimodal generation and understanding [1] - It supports various tasks including text-to-image/video generation, image editing, image translation, and image restoration [1] - This model is the first to achieve efficient end-to-end training based on the Ascend AI hardware and software platform along with the MindSpeed MM multimodal model suite [1]