plt.title('pool3 weights')
if __name__ == '__main__':
netParams = snn.params('network.yaml')
# Define the cuda device to run the code on.
device = torch.device('cuda')
# deviceIds = [2, 3]
# Create network instance.
net = Network(netParams).to(device)
# net = torch.nn.DataParallel(Network(netParams).to(device), device_ids=deviceIds)
# Create snn loss instance.
error = snn.loss(netParams, spikeLayer).to(device)
# Define optimizer module.
# optimizer = torch.optim.Adam(net.parameters(), lr = 0.01, amsgrad = True)
optimizer = optimizer.Nadam(net.parameters(), lr=0.01, amsgrad=True)
# Dataset and dataLoader instances.
trainingSet = IBMGestureDataset(
datasetPath=netParams['training']['path']['in'],
sampleFile=netParams['training']['path']['train'],
samplingTime=netParams['simulation']['Ts'],
sampleLength=netParams['simulation']['tSample'])
trainLoader = DataLoader(dataset=trainingSet,
batch_size=4,
shuffle=True,
num_workers=1)
spikeLayer1 = self.slayer.spike(self.slayer.psp(self.fc1(spikeInput)))
spikeLayer2 = self.slayer.spike(self.slayer.psp(self.fc2(spikeLayer1)))
spikeLayer3 = self.slayer.spike(self.slayer.psp(self.fc3(spikeLayer2)))
spikeLayer4 = self.slayer.spike(self.slayer.psp(self.fc4(spikeLayer3)))
spikeLayer5 = self.slayer.spike(self.slayer.psp(self.fc5(spikeLayer4)))
spikeLayer6 = self.slayer.spike(self.slayer.psp(self.fc6(spikeLayer5)))
return spikeLayer6
# Definicia zariadenia, na ktorom sa bude vykonavat trenovania
device = torch.device('cuda')
# Vytvorenie instancie triedy Network a presunutie na zariadenie
net = Network(netParams).to(device)
# Vytvorenie instancie chyby
error = snn.loss(netParams).to(device)
# Definicia optimalizacneho algoritmu a parametrov
optimizer = torch.optim.Adam(net.parameters(), lr = 0.001, amsgrad = False)
# Vytvorenie instancii dataset, zvlast pre trenovaciu a testovaciu cast
trainingSet = nmnistDataset(datasetPath =netParams['training']['path']['in'],
sampleFile =netParams['training']['path']['train'],
samplingTime=netParams['simulation']['Ts'],
sampleLength=netParams['simulation']['tSample'])
trainLoader = DataLoader(dataset=trainingSet, batch_size=8, shuffle=True, num_workers=4)
testingSet = nmnistDataset(datasetPath =netParams['training']['path']['in'],
sampleFile =netParams['training']['path']['test'],
samplingTime=netParams['simulation']['Ts'],
sampleLength=netParams['simulation']['tSample'])
model_args = {
"params": params,
"tact_input_size": 156 if args.fingers == "both" else 78,
"vis_input_size": (50, 63, 2),
"tact_output_size": 50,
"vis_output_size": 10,
"output_size": output_size,
}
device = torch.device("cuda")
writer = SummaryWriter(".")
net = model(**model_args).to(device)
if args.loss == "NumSpikes":
params["training"]["error"]["type"] = "NumSpikes"
error = snn.loss(params).to(device)
criteria = error.numSpikes
elif args.loss == "WeightedNumSpikes":
params["training"]["error"]["type"] = "WeightedNumSpikes"
error = snn.loss(params).to(device)
criteria = error.weightedNumSpikes
optimizer = torch.optim.RMSprop(net.parameters(), lr=args.lr, weight_decay=0.5)
train_dataset = ViTacDataset(path=args.data_dir,
sample_file=f"train_80_20_{args.sample_file}.txt",
output_size=output_size,
spiking=True,
mode=args.mode,
fingers=args.fingers)
train_loader = DataLoader(
if __name__ == '__main__':
netParams = snn.params('network.yaml')
# Define the cuda device to run the code on.
device = torch.device('cuda')
# deviceIds = [1, 2]
# Create network instance.
net = Network(netParams, 'emg', 'dvsCropped').to(device)
# net = torch.nn.DataParallel(Network(netParams).to(device), device_ids=deviceIds)
# Create snn loss instance.
error = snn.loss(netParams, snn.loihi).to(device)
# Define optimizer module.
optimizer = optim.Nadam(net.parameters(), lr=0.01)
# Dataset and dataLoader instances.
trainingSet = fusionDataset(
samples=np.loadtxt('train.txt').astype(int),
samplingTime=netParams['simulation']['Ts'],
sampleLength=netParams['simulation']['tSample'],
# sampleLength=2000,
)
testingSet = fusionDataset(
samples=np.loadtxt('test.txt').astype(int),
samplingTime=netParams['simulation']['Ts'],
model_args = {
"params": params,
"tact_input_size": 156,
"vis_input_size": 50 * 63,
"tact_output_size": 50,
"vis_output_size": 10,
"output_size": output_size,
}
device = torch.device("cuda")
writer = SummaryWriter(".")
net = model(**model_args).to(device)
if args.loss == "NumSpikes":
params["training"]["error"]["type"] = "NumSpikes"
error = snn.loss(params, spikeLayer).to(device)
criteria = error.numSpikes
elif args.loss == "WeightedNumSpikes":
params["training"]["error"]["type"] = "WeightedNumSpikes"
error = snn.loss(params, spikeLayer).to(device)
criteria = error.weightedNumSpikes
optimizer = torch.optim.RMSprop(net.parameters(), lr=args.lr)
train_dataset = ViTacDataset(
path=args.data_dir,
sample_file=f"train_80_20_{args.sample_file}.txt",
output_size=output_size,
spiking=True,
mode=args.mode,
loihi=True,
netParams = snn.params('network.yaml')
print('vDecay:', netParams['neuron']['vDecay'])
print('iDecay:', netParams['neuron']['iDecay'])
# Define the cuda device to run the code on.
device = torch.device('cuda:2')
# deviceIds = [1, 2]
# Create network instance.
net = Network(netParams).to(device)
# net = torch.nn.DataParallel(Network(netParams).to(device), device_ids=deviceIds)
# Create snn loss instance.
error = snn.loss(netParams, snn.loihi).to(device)
# test for whole 2 seconds of data
testLength = 200
netParamsTest = snn.params('network.yaml')
netParamsTest['training']['error']['tgtSpikeRegion']['start'] = 0
netParamsTest['training']['error']['tgtSpikeRegion']['stop'] = testLength
netParamsTest['training']['error']['tgtSpikeCount'][
False] *= testLength / netParams['simulation']['tSample']
netParamsTest['training']['error']['tgtSpikeCount'][
True] *= testLength / netParams['simulation']['tSample']
testError = snn.loss(netParamsTest, snn.loihi).to(device)
# Define optimizer module.
# optimizer = torch.optim.Adam(net.parameters(), lr = 0.01, amsgrad = True)
