Best Practice Guide on Artificial Intelligence on EuroHPC Systems

This Best Practice Guide provides practical guidance for users of EPICURE on running and scaling Artificial Intelligence (AI) workloads across EuroHPC supercomputing systems. It outlines the tools, frameworks and deployment approaches available for large-scale AI training. Next, it presents benchmark results on the scalability of different AI workloads on the different EuroHPC machines. In addition, the guide provides practical examples, job scripts and recommendations to help users efficiently deploy and scale AI applications on EuroHPC infrastructure.

This Best Practice Guide includes example job scripts and benchmark outputs collected from multiple EuroHPC machines. These resources are shared on EPICURE’s BPG-AI workloads GitLab repository, enabling users to implement and scale their workloads. Based on these results, we derive practical guidelines for running and scaling AI workloads on EuroHPC systems.