def test_SpikeSourceArray(self):
spike_times = [50.]
p = sim.Population(3, sim.SpikeSourceArray(spike_times=spike_times))
p2 = sim.Population(3, sim.Hardware_IF_cond_exp())
syn = sim.StaticSynapse(weight=0.012)
con = sim.Projection(p,
p2,
connector=sim.OneToOneConnector(),
synapse_type=syn,
receptor_type='excitatory')
spike_times_g = p.get('spike_times')
p2.record('v')
sim.run(100.0)
weights = nan_to_num(con.get('weight', format="array"))
print weights
data = p2.get_data().segments[0]
vm = data.filter(name="v")[0]
print vm
Figure(
Panel(weights,
data_labels=["ext->cell"],
line_properties=[{
'xticks': True,
'yticks': True,
'cmap': 'Greys'
}]),
Panel(vm,
ylabel="Membrane potential (mV)",
data_labels=["excitatory", "excitatory"],
line_properties=[{
'xticks': True,
'yticks': True
}]),
).save("result")
def run_experiment(self, weight_distortion=None):
import pyNN.hardware.brainscales as pynn
setup_params = dict(
useSystemSim=True,
ignoreDatabase=True,
ignoreHWParameterRanges=True,
hardware=pynn.hardwareSetup["one-hicann"],
rng_seeds=[123567],
)
if weight_distortion is not None:
setup_params['ess_params'] = {
'weightDistortion': weight_distortion
}
pynn.setup(**setup_params)
num_neurons = 2
pop = pynn.Population(num_neurons, pynn.IF_cond_exp)
# two different input spike trains, with identical rate but shifted to avoid loss in merger tree
spike_times1 = np.arange(100., 500, 10.)
spike_times2 = np.arange(105., 505, 10.)
pop_source = pynn.Population(2, pynn.SpikeSourceArray)
pop_source.tset('spike_times', [spike_times1, spike_times2])
weight = 0.1 # weight = pop[0].cell.getConductanceRange("weight")[1]
conn = pynn.OneToOneConnector(weights=weight)
proj_0 = pynn.Projection(pop_source, pop, conn)
pop.record()
pynn.run(600.)
spikes = pop.getSpikes()
spikes_nrn_0 = spikes[spikes[:, 0] == 0, 1]
spikes_nrn_1 = spikes[spikes[:, 0] == 1, 1]
pynn.end()
return len(spikes_nrn_0), len(spikes_nrn_1)
def runTest(self):
import pyNN.hardware.brainscales as pynn
from pyNN.hardware.brainscales import mapper
import os
place = mapper.place()
current_directory = os.path.dirname(os.path.realpath(__file__))
pynn.setup(timestep=0.1,
useSystemSim = True,
algoinitFilename= current_directory + "/algoinit_l1_delay.pl",
mappingStrategy="user",
loglevel=2,
hardware=[dict(setup="wafer", wafer_id=0,hicannIndices=[279,280])],
hardwareNeuronSize=1,
logfile='logfile.txt',
speedupFactor=10000,
rng_seeds=[123567],
ignoreHWParameterRanges=True,
ignoreDatabase=True,
tempFolder="debug_l1_delay",
)
LIF_nrn_params = {
'cm': 1.0,
'e_rev_E': 0.0,
'e_rev_I': -70.0,
'i_offset': 0.0,
'tau_m': 20.0,
'tau_refrac': 5., # high value to avoid 2nd spike
'tau_syn_E': 5.0,
'tau_syn_I': 5.0,
'v_reset': -65.0,
'v_rest': -65.0,
'v_thresh': -60.0} # low value for 1 spike
num_pe = 8 # number of priority encoders
pop_1 = pynn.Population(num_pe, pynn.IF_cond_exp, LIF_nrn_params)
pop_2 = pynn.Population(num_pe, pynn.IF_cond_exp, LIF_nrn_params)
pop_source = pynn.Population(num_pe, pynn.SpikeSourceArray)
spike_times = [[i*10. + 10.] for i in range(num_pe)]
pop_source.tset('spike_times',spike_times)
weight = 0.05
conn = pynn.OneToOneConnector(weight,0.1)
proj_0=pynn.Projection(pop_source,pop_1, conn)
proj_1=pynn.Projection(pop_1,pop_2, conn)
for i in range(num_pe):
place.to(pop_1[i], hicann=279, neuron=32*i)
place.to(pop_2[i], hicann=279, neuron=32*i +1)
place.commit()
pop_1.record()
pop_2.record()
pynn.run(100.)
spikes_1 = pop_1.getSpikes()
spikes_2 = pop_2.getSpikes()
self.assertEqual(8,len(spikes_1))
self.assertEqual(8,len(spikes_2))
# sort spikes by ID:
spikes_1.sort()
spikes_2.sort()
# spiketimes only
spikes_1 = spikes_1[:,(1,)]
spikes_2 = spikes_2[:,(1,)]
delays = spikes_2 - spikes_1
# delay pe 1 > 0, 2, 4, 7
self.assertTrue(delays[1] > delays[0])
self.assertTrue(delays[1] > delays[2])
self.assertTrue(delays[1] > delays[4])
self.assertTrue(delays[1] > delays[7])
# delay pe 6 > 0, 2, 4, 7
self.assertTrue(delays[6] > delays[0])
self.assertTrue(delays[6] > delays[2])
self.assertTrue(delays[6] > delays[4])
self.assertTrue(delays[6] > delays[7])
# delay pe 5 > than everything expect 3
self.assertTrue(delays[5] > delays[0])
self.assertTrue(delays[5] > delays[1])
self.assertTrue(delays[5] > delays[2])
self.assertTrue(delays[5] > delays[4])
self.assertTrue(delays[5] > delays[6])
self.assertTrue(delays[5] > delays[7])
# delay pe 3 > than everything
self.assertTrue(delays[3] > delays[0])
self.assertTrue(delays[3] > delays[1])
self.assertTrue(delays[3] > delays[2])
self.assertTrue(delays[3] > delays[4])
self.assertTrue(delays[3] > delays[5])
self.assertTrue(delays[3] > delays[6])
self.assertTrue(delays[3] > delays[7])
def runTest(self):
import numpy
backend = "hardware.brainscales"
import pyNN.hardware.brainscales as pynn
# set-up the simulator
pynn.setup(
timestep=0.1,
useSystemSim=True,
speedupFactor=8000,
ignoreDatabase=True,
ignoreHWParameterRanges=True,
hardware=pynn.hardwareSetup['one-hicann'],
)
neuron_count = 1 # size of the Population we will create
# Set the neuron model class
neuron_model = pynn.EIF_cond_exp_isfa_ista # an Adaptive Exponential I&F Neuron
neuron_parameters = {
'a' : 10.0, # adaptation variable a in nS
'b' : 0.0805, # adaptation variable b in pA
'cm' : 0.281, # membrane capacitance nF
'delta_T' : 2.0, # delta_T fom Adex mod in mV, determines the sharpness of spike initiation
'e_rev_E' : 0.0, # excitatory reversal potential in mV
'e_rev_I' : -100.0, # inhibitory reversal potential in mV
'i_offset' : 0.0, # offset current
'tau_m' : 9.3667, # membrane time constant
'tau_refrac' : 8.0, # absolute refractory period
'tau_syn_E' : 5.0, # excitatory synaptic time constant
'tau_syn_I' : 5.0, # inhibitory synaptic time constant
'tau_w' : 144.0, # adaptation time constant
'v_reset' : -70.6, # reset potential in mV
'v_rest' : -70.6, # resting potential in mV
'v_spike' : -40.0, # spike detection voltage in mV
'v_thresh' : -66.9, # spike initiaton threshold voltage in mV
}
# We create a Population with 1 neuron of our
N1 = pynn.Population( size = 1, cellclass = neuron_model, cellparams = neuron_parameters)
N1.initialize('v',neuron_parameters['v_rest'])
spiketimes = numpy.arange(100.,800.,0.5)
S1 = pynn.Population(1,pynn.SpikeSourceArray, cellparams = {'spike_times':spiketimes})
pynn.Projection(S1,N1,pynn.OneToOneConnector(weights=0.0123938304114), target='inhibitory') # max weight in hardware
# record the spikes of all neurons of the population
N1.record()
# run the simulation for 500 ms
pynn.run(1000.)
num_spikes = 0
# After the simulation, we get Spikes and save the voltage trace to a file
spike_times = N1.getSpikes()
for pair in spike_times:
print "Neuron ", int(pair[0]), " spiked at ", pair[1]
num_spikes += 1
pynn.end()
self.assertEqual(int(num_spikes),1,'The Neuron spiked more than once, i.e. there was a rebound burst, and not a single rebound spike as it should be')
