Support queries involving Python are becoming more and more common in HPC. Python can feature within traditional HPC workflows and is particularly popular for machine learning tasks. However, those Python packages involved with inter-process communications (mpi4py) and low-level computation (pycuda/cupy) must be linked to code libraries built specifically for the host system. It is more convenient therefore to maintain such packages within a centralized location, and, at the same time, provide users with a straightforward way to access those packages from within their own local Python environments.
This talk explains how this can be accomplished and then proceeds through several use cases implemented on the two main HPC systems delivered by EPCC, namely ARCHER2 and Cirrus.
This online session is open to all. It will use the Blackboard Collaborate platform.