Ds Scholarship

Large AI Language Model and AI in Scientific Research

Beijing, January 17, 2022–(BUSINESS WIRE)–The preliminary round of the 2022 Student Supercomputer Challenge (ASC22) kicked off on January 5, 2022. More than 300 teams around the world are competing in challenging missions such as the Yuan 1.0 large AI language model and the 2020 DeePMD-winning Gordon Bell award-winning kit.

Among the participants are championship teams from previous SC, ISC and ASC student group competitions, including Tsinghua University (ISC21 and SC21 champion), Jinan University (ASC20-21 on-site champion, ISC21 third place), National Tsinghua University ( ASC20-21 Virtual Champion), Sun Yat-sen University (SC21- IndySCC Champion), ShanghaiTech University (runner-up in SC21), Southern University of Science and Technology (LINPACK winner in SC21), ETH Zurich (3rd place) place in SC21- IndySCC). Other strong contenders include teams from the University of Science and Technology of China, Texas A&M University, Ural Federal University, Monash University and Ifte University.

In the preliminary round of ASC22, all teams have to tackle two highly advanced tasks: Yuan1.0 – a great AI language model, and an excellent suite of AI with an advanced science application – the DeePMD-kit.

Launched in 2021, Yuan 1.0 is a record-breaking mega form of Chinese with 247.5 billion parameters. It has shown promising performance in various language processing applications. Training such a massive model of AI requires massive computing tasks for parallel computing, offering student teams a real-life test of their skills in improving supercomputing applications. Competitors have to build a scale model of the yuan with 4.7 billion parameters based on a 100GB high-quality Chinese dataset provided by the ASC22. Without any reference code being provided to design the model, teams need to create the model, identify optimal training strategies, and achieve optimal performance.

DeePMD-kit, a program designed for machine learning-based molecular dynamics, offers an excellent opportunity for students to explore the possibilities of combining artificial intelligence with science, which has become one of the most promising synergies of breakthrough innovations in the field. Advances in High Performance Computing (HPC) have unleashed the huge potential of ML-based molecular dynamics. The 2020 Gordon Bell Award winner DeePMD has achieved double-precision performance of 91 PFLOPS and time to resolution 1,000 times faster than the previous baseline performance in the Summit supercomputer, thus redefining the science, AI, and HPC integration model. Student teams will have hands-on experience building a model using the DeePMD suite, and optimizing the HPC system to deliver maximum training speed.

The ASC22 Opening Ceremony will be held in Beijing on January 19, 2022. Important information such as the host of the competition finals will be announced at this event. Then, all teams will have the opportunity to participate in the online ASC22 bootcamp from January 20 to January 21, 2022. HPC and AI experts from Intel, NVIDIA, AWS Cloud, DP Technology and State Key Laboratory of High-end Server & Storage Technology will Technology by lecturing teams on how to prepare for competition by advising on competition rules, designing and optimizing HPC on-premises and cloud suite, etc. Winners from previous seasons will also share their experiences during the bootcamp.

About ASC

The ASC Student Supercomputer Challenge is sponsored and organized by the Asian Supercomputer Society and supported by Asian, European and American experts and institutions. The main objectives of ASC are to encourage the exchange and training of young supercomputing talents from all over the world, improve supercomputing applications and research and development capacity, promote the development of supercomputing, and promote technical and industrial innovation. ASC’s first Student Supercomputer Challenge was held in 2012 and has since drawn more than 10,000 undergraduate students from all over the world. Learn more about ASC at http://www.asc-events.net/.

View the source version on businesswire.com: https://www.businesswire.com/news/home/20220117005139/ar/


Fiona Leo


Please enter your comment!
Please enter your name here