EBRAINS Framework Unifies HPC and Neuromorphic Computing for Seamless Workflows

Researchers from EBRAINS have unveiled a unified, cloud-based workflow that bridges von-Neumann supercomputers and neuromorphic accelerators, addressing a long-standing challenge in computational neuroscience. The framework, orchestrated via EBRAINS JupyterLab, allows users to execute spiking neural networks on HPC sites like JUSUF and Galileo100 using PyUNICORE, and on the SpiNNaker-1 neuromorphic system via the Neuromorphic Computing Platform Interface—all with a single federated identity. This eliminates the need for fragmented, site-specific pipelines that force researchers to maintain separate code and configurations for each architecture.

To guarantee reproducibility across sites, the system leverages a zero-installation execution mode that dynamically pulls PMIx-aware Apptainer containers to compute nodes, mitigating software version drift. A key innovation is the use of NESTML, a domain-specific language that allows custom neuron models to be written once and automatically compiled for either the NEST (C++) or sPyNNaker backends. The workflow was validated with a balanced random network case study, demonstrating true model-level portability. This end-to-end solution highlights how containerization and DSLs can unify heterogeneous computing resources for high-performance neuroscience workflows.