Hi Russell,

Being able to easily build and simulate SONATA networks with NEST is hopefully imminent. We currently have a pull request under review which will add native NEST support for networks described by the SONATA format, see https://github.com/nest/nest-simulator/pull/2595.

With our implementation, 64 GB of memory is more than sufficient for building and simulating the entire point neuron version of the Allen V1 model.

Regarding a downscaled version of the V1 model, this cannot be achieved solely through the JSON configuration file for this particular model. This is due to all layers (V1 nodes) and associated edges being contained in single HDF5 files.

In order to use a downscaled V1 model in the context of the upcoming NEST functionality, you would actually need to construct a new set of SONATA files that contain just the nodes and associated edges you are interested in. That is, you have to parse each nodes and edges .h5 and .csv file, extract the relevant entries, create new files and store the relevant entries in correctly named datasets and in the correct order.

Unfortunately, I do not know of any examples that can guide you in this endeavour. You could try to reach out to the SONATA developers here https://github.com/AllenInstitute/sonata. They might have some ideas or even readily available examples.

Hope this helps!

Best regards,

Nicolai