Updating Extra Global Parameters at runtime

[why9928]

# flat_neuron_spike_times is input time
stim_ini = {"startSpike": np.append(0, num_spikes_per_neuron_cumulative[:-1]),
            "endSpike": num_spikes_per_neuron_cumulative}

P = model.add_neuron_population("Stim", new_resolution[0] * new_resolution[1], "SpikeSourceArray", {}, stim_ini)

P.set_extra_global_param("spikeTimes", flat_neuron_spike_times)

model.add_synapse_population(
    "p_bsynapse", "SPARSE_GLOBALG", NO_DELAY, P, B_pop,
    "StaticPulse", {}, {"g": 1.0}, {}, {},
    "DeltaCurr", {}, {},
    init_connectivity("OneToOne", {}))

P.spike_recording_enabled = True
B_pop.spike_recording_enabled = True

model.build()
model.load(num_recording_timesteps=resent_timesteps)

while model.t < sim_time:
    model.step_time()

# flat_neuron_spike_times2 is other input time data
P.set_extra_global_param("spikeTimes", flat_neuron_spike_times2)
P.push_extra_global_param_to_device("spikeTimes")


How does Genn update the data entered? I want to update P's data every sim_time, i.e. update the input

[why9928]

How to enter a batch of data into the model every 200ms, and then I do the data processing operations without stopping the model

[neworderofjamie]

So what you need to do is:

1. When you first initialize the extra global parameter with P.set_extra_global_param("spikeTimes", flat_neuron_spike_times), you need to make it large enough for your largest spike array (I tend to use np.empty(max_size) for this.
2. Every 200ms, update the extra global parameters through its 'view' at P.extra_global_params["spikeTimes"].view[:len(spikes)] = spikes
3. Call P.push_extra_global_param_to_device("spikeTimes") are you are already doing

[why9928]

thank you for your reply