The largest cloud based research intelligent computing dedicated platform for universities in China has officially launched CFFF | Platform | Computing
On June 27th, the largest cloud based research intelligent computing platform for universities in China, Fudan University CFFF "Qiewen" No.1 and "Jinsi" No.1, were officially launched. Fudan University has also signed strategic cooperation agreements with China Telecom Group Co., Ltd. and Alibaba, and launched the first World Science and Intelligence Competition and held the first Science and Intelligence High end Forum.
The intelligent computing platform CFFF, jointly deployed and constructed by Fudan University and China Telecom, includes the "Qiewen-1" scientific intelligent computing cluster and the "Jinsi-1" dedicated high-performance computing cluster. The total scale of CFFF computing power has reached 28 PFlop/s. China Telecom connects the Shanghai Fudan campus computer room with the Alibaba Cloud Ulanqab Super Data Center 1500 kilometers away through the 100GOTN high-speed dedicated line, ensuring high-speed data exchange on and off campus. In the past, it used to take two weeks for PB level scientific research data to be transmitted from Fudan University to the Western Data Center. Now, it can be completed on the same day, truly achieving "East to West Calculation".
This project is also the first large-scale public computing power project of Shanghai Telecom, consolidating the technical foundation of Shanghai Telecom's public service computing power project. In the future, it will continue to expand its computing power scale and open up to other research institutions in Shanghai.
On that day, the first World Science Intelligence Competition was officially launched. The competition is hosted by Fudan University and collaborates with Alibaba Cloud Tianchi Platform, China Telecom, and the Cloud Institute of the Chinese Academy of Information and Communications Technology to set up five major tracks: life science, atmospheric science, materials science, fluid mechanics, and quantum chemistry. It is expected to recruit 5000 innovation teams from five continents, and use the CFFF platform as a support to embark on a new exploration journey of basic scientific research, promoting the early discovery, cultivation, and early transformation of scientific and intelligent innovative talents.
At the meeting, the competition questions for the five major tracks were announced, including methylation biological aging clock, AI mid-term weather prediction in the East China region, estimation of MOF material synthesis conditions, flow problem solving based on NS equation, and molecular property prediction. The competition sets up a prize pool of up to 1 million yuan, and the winners of each track will also receive computing power support from the CFFF platform, as well as opportunities for scientific research cooperation with the Artificial Intelligence Innovation and Industry Research Institute of Fudan University and various basic disciplines.
At present, the first scientific research achievement on the CFFF platform has been born. The team led by Li Hao from the Institute of Artificial Intelligence Innovation and Industry Research at Fudan University recently released a large model for medium and short term weather forecasting with a parameter size of 4.5 billion. The prediction performance has reached the industry recognized average level of the ECMWF set for the first time on public datasets, and the prediction speed has been reduced from the original hourly level to 3 seconds. In just one day, a large model with 4.5 billion parameters can complete training, which is difficult for traditional computing platforms to achieve.