/PRCWT/Qingdao, September 04, 2023——August 23-26, 2023, hosted by China Computer Society, The 19th National High Performance Computing Annual Conference (CCF HPC China 2023), co-sponsored by China Computer Society High Performance Computing Committee, Ocean University of China, Qingdao Ocean Science and Technology Center and Qilu University of Technology (Shandong Academy of Sciences), was held in Qingdao Hongdao International Convention and Exhibition Center.
This conference gathered 9 Chinese and foreign academicians, 11 invited reports, 6 industry reports, 46 theme forums, more than 30 special activities, 10,000 computing power exhibition hall, a total of more than 3,000 people gathered. Beijing Super Cloud Computing Center (hereinafter referred to as "Beijing Supercomputer") was invited to participate in the exhibition as a platinum partner, and won the AI large model computing platform solution award.
Facilitates large model training and reasoning
"Supercomputing architecture" speeds up training of large models
With the popularity of ChatGPT, it has aroused widespread concern at home and abroad, a number of large model applications have emerged in China, and the demand for high-end computing resources has become increasingly strong, forming a huge gap between demand and supply. As a special guest of the conference, Wu Di, general manager of Beijing Supercomputer, made a theme report entitled "AI intelligent computing Cloud helps large model training and reasoning".
Wu Di pointed out in the report: "Beijing super cloud computing Center uses the supercomputing architecture model to build intelligent computing resources, and integrates the resources of major computing power centers distributed in China, which can maximize the use of existing resources, reduce the loss caused by the waste of idle resources, and also effectively meet the needs of high-end computing power for AI large model training." Due to the characteristics of heterogeneous computing and massively parallel computing, it is an ideal choice for large model training of artificial intelligence. Supercomputing architectures, represented by supercomputers, can provide more comprehensive and efficient parallel computing capabilities, provide stronger support for large model training processing, achieve more efficient model training, faster training convergence, and better prediction accuracy."
Build a solid digital base to strengthen the computing power
Beijing Supercomputer won the AI Large Model Computing Power Platform Solution Award
At the award dinner, the supercomputing architecture computing power service capability based on Beijing Supercomputer, while supporting the large-scale computing power required for large-scale model training, can improve data storage and processing capabilities, accelerate the efficiency of AI large-scale model training, and meet the growing demand for large-scale model computing infrastructure. Beijing Supercomputer won the AI Large Model Computing Power Platform Solution Award.
In 2023, Beijing Supercomputer will focus on the layout of artificial intelligence application market, become the first batch of computing power partners in Beijing, and provide GPU computing power resources based on supercomputing architecture required for large model pre-training for well-known domestic AI enterprises such as Zhiyuan Research Institute and Zhipu ·AI.
Wise future
The Development and Application Forum of supercomputing Enabling large models was successfully held
"The integration of supercomputing and intelligent computing is an inevitable trend in the future, and the development of supercomputing will bring more possibilities to intelligent computing." Chi Xuebin, chief engineer of the Computer Network Information Center of the Chinese Academy of Sciences, opened the prelude to the development and application forum of supercomputing enabling large models.
At the roundtable discussion, in view of how to deal with the bottleneck of computing power in the development of large models, the participating experts pointed out that by promoting the docking of supply and demand, the transformation of computing power network can be accelerated, and the centralized management and efficient utilization of computing power can be realized. Optimize algorithms and architectures, make full use of computing resources through distributed computing, and improve computing speed and efficiency; In addition, experts called on all parties to work together to promote the innovative research and development of domestic chip technology and the construction of application ecology.
Win-win cooperation
Beijing supercomputer signed a blockbuster contract with Wenyin Internet and North Taitian Yuan
Beijing Supercomputer and Wenyin Internet reached a strategic cooperation, based on the advantages of Beijing Supercomputer GPU computing resources and supercomputing architecture service capabilities, as well as Wenyin Internet self-developed vertical field large model and years of financial industry scenario case accumulation, plan to jointly build a large model open innovation computing platform, provide more value-added services for industry users, and jointly create a good industry ecology. Promote business innovation and technological upgrading.
In the same period, Beijing Supercomputer signed a strategic cooperation agreement with Beitai Tianyuan, a pioneer of domestic general-purpose scientific computing software. Based on Beijing Supercomputer's computing power public service platform and Beitai Tianyuan's deep research and development and technical strength, the two sides will combine the advantages of scientific computing software and supercomputer, promote the ecological integration and technological integration of the two sides, and constantly expand the breadth and depth of strategic cooperation. Achieve win-win cooperation.
Wanm exhibition hall Beijing supercomputer booth is favored
As a platinum partner, Beijing supercomputer booth in the form of a special appearance in the Wanmi exhibition hall, the booth with "AI" as the design core, full of science and technology, and very visual impact, giant AI light box and "supercomputer architecture large model computing power leader" Slogan highlights the determination and strength of Beijing supercomputer to deeply farm large model computing power market.
During the conference, the "Super Face to face" activity held at the Beijing Supercomputer booth was welcomed by the audience, and the audience continued to stream, and the wonderful place won bursts of applause.
On the booth, the AI intelligent computing Cloud commemorative medallion designed by Beijing Supercomputer is composed of four commemorative medallions of "AI", "large model", "computing power" and "Leader", and a sterling silver "AI" commemorative medallion is specially customized, and five constitute a set. Once it appeared, it detonated the whole audience, and the audience gathered a set of commemorative MEDALS through participating in the interaction.
As one of the three most influential supercomputing events in the world's high-performance computing field, CCF HPC China is also the largest, most influential and highest academic supercomputing event held in China, which has played a leading and promoting role in the development of domestic supercomputing. Beijing Supercomputer participated in the conference for three consecutive years, continued to deepen communication and cooperation with the industry, promoted the integration and development of supercomputing and intelligent computing, provided rich, safe and reliable computing resources for large model enterprises and scientific research users, and committed to becoming a leader in large model computing power of supercomputing architecture.
