Looks like i have solved this issue now. 

For those interested in the solution, i have switched from using the efficient method for splitting a simulation into multiple intervals described in the docs, which has the clean up functions outside of the loop, to just using nest.Simulate() in the loop so the clean up functions are called each time.

Cheers,
Rachael


From: Rachael Stentiford <rachael.stentiford@brl.ac.uk>
Sent: Tuesday, April 12, 2022 6:32 PM
To: users@nest-simulator.org <users@nest-simulator.org>
Subject: [NEST Users] Kernel death when recording from 300+ cells for 2mins+
 
Dear all,

I hope you are doing well. I am having an issue using pyNEST which i haven't come across before and have been unable to find answer to elsewhere. I am working with Python 3.8 and Nest 2.18 (which is the version currently used by the Neurorobotics platform).

I have a network of 6 x 360 neurons, and i am trying to record spikes from 360 of these. However, I am finding that if i try to simulate for more than ~2 mins i run into problems. In jupyter notebooks I get a kernel death error, but even just running a standard python script NEST dies. I can simulate for longer if i only record from a single neuron.

Previously I have used networks of 4 x 180 neurons and recorded from 180 for 10+ minutes without an issue. 

I have tried splitting up the simulation into multiple intervals, as explained in the docs, and saving spikes in between. This would mean there are fewer spikes to save each time, but this doesn't appear to help.
Ideally i would like to record from 360 cells for around 30 mins. Have any of you come across a similar issue in the past, or have any suggestions for how I could fix this issue? 

Thanks for your help!

Rachael

-------------------------
Dr Rachael Stentiford (she/her)
Research Associate in Neurobotics

Bristol Robotics Laboratory

University of the West of England

Bristol

BS16 1QY