Is there a way to clean spike storage without restarting simulation?

[andreARibe]

Hello,

I am running a large simulation, where I need to record, for an extended number of time-steps, the spikes spikes of every neuron in the simulation. Naturally, I came to a point where I use all the GPUs resources.

Is it possible to pull spikes from the simulation, store them, clean the buffer, and continue?

I would be looking to something like this:

` start = time.time()
while self.model.timestep < timesteps:
self.model.step_time()

        if self.model.timestep % buffer_t == 0:
        
            self.model.pull_recording_buffers_from_device()
            # store ()
            # clean device buffers()

`

Thank you!

[neworderofjamie]

as long as you have loaded the model with the buffer size set to match buffer_t i.e.:

self.model.load(num_recording_timesteps=buffer_t)

this code will already work the way you want. You just need to make a copy of the data after calling pull_recording_buffers_from_device() e.g.

spike_times, spike_ids = genn_pop.spike_recording_data[0]
self.spike_times.append(spike_times)
self.spike_ids.append(spike_ids)