def __init__(self):
print("Model define Start ! ")
folder = "models/"
nfilters_embeded = int(64)
nlayers_embeded = int(4)
nfilters_cloud = int(64)
nlayers_cloud = int(4)
branchweight = float(0.1)
lr = numpy.float64(0.001)
nepochs = int(80)
ent_T = numpy.float64(0.4)
name = str("cifar10_64f_4l")
input_dims = 3
output_dims = 10
model = Sequential()
for i in range(nlayers_embeded):
if i == 0:
nfilters = input_dims
model.add(
ConvPoolBNBST(nfilters, nfilters_embeded, 3, 1, 1, 3, 1,
1))
else:
nfilters = nfilters_embeded
model.add(
BinaryConvPoolBNBST(nfilters, nfilters_embeded, 3, 1, 1, 3,
1, 1))
branch = Sequential()
branch.add(BinaryLinearBNSoftmax(None, output_dims))
model.add(branch)
for i in range(nlayers_cloud):
if i == 0:
nfilters = nfilters_embeded
else:
nfilters = nfilters_cloud
model.add(Convolution2D(nfilters, nfilters_cloud, 3, 1, 1))
model.add(BatchNormalization(nfilters_cloud))
model.add(Activation('relu'))
model.add(max_pooling_2d(3, 1, 1))
model.add(Linear(None, output_dims))
model.build()
chain = Chain(branchweight=branchweight, ent_T=ent_T)
chain.add_sequence(model)
chain.setup_optimizers('adam', lr)
print("Model define Over ! ")
##################################Skip the training ##################################################
##################################Load the model after training by other code #########################
print("Loading model Start ! ")
#name = self.get_name(**kwargs)
#path = '{}/{}'.format(self.folder,name)
#epoch = int(kwargs.get("nepochs",2))
#fn = "{}/chain_snapshot_epoch_{:06}".format(path,epoch)
#cf = os.path.dirname(os.path.abspath(__file__))
fn = "models/test_cifar10/cifar10_64f_4l/epoch_000080"
#ph = cf + fn
#print "path %s" % (ph)
S.load_npz(fn, chain)
self.chain = chain
print("Loading model Over ! ")
nfilters_cloud = int(64)
nlayers_cloud = int(4)
branchweight = float(0.1)
lr = numpy.float64(0.001)
nepochs = int(80)
ent_T = numpy.float64(0.4)
name = str("cifar10_64f_4l")
input_dims = 3
output_dims = 10
model = Sequential()
for i in range(nlayers_embeded):
if i == 0:
nfilters = input_dims
model.add(ConvPoolBNBST(nfilters, nfilters_embeded, 3, 1, 1, 3, 1, 1))
else:
nfilters = nfilters_embeded
model.add(
BinaryConvPoolBNBST(nfilters, nfilters_embeded, 3, 1, 1, 3, 1, 1))
branch = Sequential()
branch.add(BinaryLinearBNSoftmax(None, output_dims))
model.add(branch)
for i in range(nlayers_cloud):
if i == 0:
nfilters = nfilters_embeded
else:
nfilters = nfilters_cloud
model.add(Convolution2D(nfilters, nfilters_cloud, 3, 1, 1))
