Using containers to simplify ML training on Berzelius and other supercomputers: beginner-friendly introduction

Abstract: A common approach to train machine learning models is to first create a prototype using a small dataset on a local machine to verify that it works and thereafter use a large scale compute infrastructure such as Berzelius for the full-scale training. One of the challenges with this approach however is incompatible systems in terms of differences in available software packages, versions, etc. An effective way to solve this issue is to use a container solution. Using a container environment allows a highly portable workflow and reproducible results between systems as diverse as a laptop, Berzelius or EuroHPC resources such as LUMI for instance. During this beginner-friendly event, we will introduce and demonstrate how to work with containers on Berzelius (Apptainer and Enroot) using an example from life sciences, starting from raw data and finishing with a trained model. During the Q&A session, the Berzelius life science support team will answer your questions.