zidongtaichu: OPT（Omni-Perception Pre-Trainer）

Zidong Taichu is a collaboration between Huawei Ascend Computing and Wuhan Institute of Artificial Intelligence and Institute of Automation, Chinese Academy of Sciences, based on the full stack of Ascend Ascend's software and hardware The Taichu series of models is an important achievement of CAS Automation and Huawei in exploring the path of general artificial intelligence. In September 2021, the world's first 100 billion model of three modes of graphic, text and audio was released, and in September 2022, the industry's largest Chinese multimodal dataset, Zidong, was released Taisu, Zidong The release of Taichu will change the current AI R&D paradigm of single model corresponding to single task, realize multimodal unified semantic expression, and have huge market value in commercial applications such as multimodal content understanding, search, recommendation and question answering, speech recognition and synthesis, human-computer interaction and unmanned driving.