The National Computational Infrastructure (NCI) in collaboration with HPC-AI Advisory Council and the National Supercomputing Centre (NSCC) Singapore, proudly supported the seventh Asia-Pacific High-Performance Computing and Artificial Intelligence (APAC HPC-AI) Competition, continuing its commitment to fostering excellence in high-performance computing (HPC) and AI. This year, the Monash University undergraduate team, DeepNeuron, once again showcased their exceptional skills, securing a Merit Award for the second consecutive year at this prestigious competition.
About the APAC HPC-AI Competition
The APAC HPC-AI Competition brings together student teams from across the Asia-Pacific region to tackle complex computational challenges using cutting-edge HPC and AI techniques. The competition aims to equip students with essential technical expertise, problem-solving skills, and experience working on real-world computing problems. In 2024, 41 teams representing 23 universities and research institutions across nine countries participated, with 21 teams advancing to the final stage to present their work.
The competition featured two main challenges this year: HOOMD-blue, a particle simulation toolkit, and Llama 2, a family of pre-trained and fine-tuned large language models (LLMs) from Meta AI. NCI provided critical access to its high-performance network and GPU facilities, empowering students to accelerate their simulations and AI models.
Monash University’s Winning Project
The Monash DeepNeuron team worked on two standout projects in this year’s competition:
- HOOMD-blue Simulations: The team focused on optimizing HOOMD-blue simulations, which are crucial for modeling complex particle systems. This has significant applications in areas like materials science, where particle-level simulation plays a role in advancing technology and research.
- Llama-3 Development: The team also contributed to the development of Llama-3, a state-of-the-art large language model leveraging an advanced AI cluster for distributed inference. Their work demonstrated innovations in computational efficiency and scalability, especially in the context of fine-tuning Llama 2 models to meet specific research needs.
The winning team members consist of Michael Xue, David Le, Daniel Arnould, Kevin Duignan and Martin Carroll, with Tyler Swann providing valuable contributions as a former team lead. The team was supervised and guided by Simon Michnowicz, Senior HPC Applications Specialist from Monash eResearch.
The Role of NCI in Supporting the Competition
As a key sponsor of the APAC HPC-AI Competition, NCI provided vital resources, technical support, and expertise to help student teams navigate the demanding challenges of the competition. NCI’s involvement was critical in allowing the team to leverage the NCI NF high-performance network (Infiniband) to accelerate OpenMPI communication within HOOMD-blue simulations. Additionally, NCI’s GPU facilities were instrumental in the Llama 2 fine-tuning, enabling the team to scale their AI models efficiently.
NCI’s involvement in the competition aligns with its broader mission to advance computational research and develop talent in HPC and AI. Through initiatives like this, NCI helps students gain hands-on experience, refine their technical skills, and prepare for impactful careers in science and technology.
Overcoming Challenges and Mastering New Technologies
For the Monash team, the competition was an opportunity to face real-world challenges in high-performance computing and AI. The open-ended nature of the competition meant there was no fixed endpoint, pushing the team to constantly refine their performance and surpass their previous benchmarks. This led to challenges in setting performance goals and determining when their work was "good enough," as they kept striving for improvement.
The steep learning curve was another major hurdle. The team had to master a variety of complex technologies and methodologies that were not part of their regular curriculum. These included learning how to use NCI’s high-performance clusters, mastering OpenMPI programming, and becoming proficient in GPU programming for AI model fine-tuning.
However, with the help of weekly tutorials by APAC Supervisor Pengzhi Zhu and local tutorials provided by Monash eResearch, the team successfully navigated these hurdles.
In addition to technical skills, the team also developed invaluable soft skills in collaboration, problem-solving, and time management. Juggling university assignments alongside the competition workload was another challenge that the team managed successfully, demonstrating their ability to work under pressure and meet deadlines.
Impact and Future Prospects
For the Monash students, winning a Merit Award at the APAC HPC-AI Competition is a testament to their technical expertise and a stepping stone to future opportunities in HPC and AI research and industry. Competitions like this provide invaluable exposure to real-world scenarios, opening doors to research collaborations, internships, and career prospects.
Their success highlights the importance of such competitions in fostering innovation and skill development in HPC and AI. The experience gave the team the confidence and tools to tackle future challenges, positioning them for success in future contests, research projects, or industry roles.
Additionally, the students gained essential skills that are not typically taught in undergraduate courses, such as working with advanced computational facilities, managing large-scale projects, and collaborating with field experts. These skills are crucial for their future careers in AI and HPC, enabling them to enter the industry with confidence.
NCI is proud to continue supporting initiatives that invest in emerging talent and drive the boundaries of computational science forward.
For more details on the competition results, visit the official announcement here.
