Documentation of external models (Jessica, Renan, Gorka, Markus, Johanna, Tom)
- microcircuit extraction into separate repo
- would maybe allow inporting MC in multi-area model, i.e. really use MC as building block
- also MC could be conctriucted of components (balanced network per layer)
Refactoring of Node and its derived classes (Pooja, Catherine, Jan, Charl, Dennis, Hans Ekkehard, Susanne)
- astrocyte model now derived directly from Node instead of SPNode
- discussion of possibilities to use components for different archivers
- TODO: investigate variadic templated as an option to implement a component-based solution
- TODO: discuss also Devices
NEST on social media (Anno, Jessica, Sebastian, Charl, Johanna, NEST board)
- NEST Initiative has a bluesky account (confirmation still needed)
- intro post as soon as official handle is available
JEDI/JUPITER (Anno, Dennis, Jan, Jose, Markus, Susanne)
- student assistant Alp onboarded for usage of beNNch
- thread-scaling (single MPI) looks good on Hambach 52 but JEDI
- microcircuit benchmarks on JEDI show "dark time"
- TODO: thread-sppecific detailed times implementation by Jan needs to be used in further benchmarks
SIONlib (Junxian, Aravind, Jan, Dennis, Susanne)
- for now use get_connections() to also trigger writing of source, target, weight, delay
- problem: retrieval of synaptic weights
- Jan shared a branch that implements connection dumping and get_weight() for static synapse
morph (Melissa, Markus, Susanne)
- logging of cycles of simulations
- pertubations for original multi-area model on JURECA but not on HAWK (using ci-beNNch)
- however, manual runs on JURECA show fewer pertubations
- TODO: use gather meta data in ci-beNNch
cMAM build-time improvements (Renan, Markus, Hans Ekkehard, Catherine)
- continued work from last hackathon
- resolved problems in using Pandas library
- PyNEST-NG code also seems to improve performance
- tests with EXABRAINPREP code revealed a few issues related tom memory usage and reproducibility
- several calls to build network (but for several clusters) worked
eprop (Agnes, Jesus, Hans Ekkehard, Markus, Susanne)
- Jesus will present at IAS-6 on May 11
- bio-eprop merged
- now working on extracting on Graz-models into a repo
- implementation of pure-activation spikes concept to ensure clearing of postsynaptic history
PyNEST-NG porting of MPI tests
- moved a few more tests from SLI to Py
cibennch (Anno, Dennis, Jan)
- disscussed concept for beNNch as a generic tool
- final decisions regarding manuscript
NESTML (Charl, Pooja)
- improvements based on user feedback (tutorials and mailinglist)
- related to PyNN: first draft PR for compatibility (part of PyNN repo)
- soon new release 8.0.1
- reviewing and bug fixes
- tests for adding numeric solver for non-linear synapses