We are organizing a Heat workshop this November at Jülich Supercomputing Centre and are looking forward to your participation! This will happen immediately after the tutorial (Don’t forget to register for the tutorial as well).

The Release 1.3.0 is out. We now support PyTorch 2 and computations on AMD GPUs. Moreover, we have added an implementation of distributed Singular Value Decomposition and a module for sparse DNDarrays. Finally, we make a docker image of Heat available.

We have implemented an algorithm for computing an approximate, truncated SVD/PCA in Heat; see heat.linalg.hsvd_rank and heat.linalg.hsvd_rtol. What this actually is, why it is helpful, and how it works, will be explained below.

On Feb 21st we learned that our Organisation has not been selected this year. Congratulations to the GSoC 2023 Mentoring Orgs!

The Bug Fix release 1.2.2 is out. It now supports newer OpenMPI versions.