def __init__(self, in_planes, planes, stride=1, option='A'):
super(BasicBlock_1w1a, self).__init__()
self.conv1 = ir_1w1a.IRConv2d(in_planes,
planes,
kernel_size=3,
stride=stride,
padding=1,
bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.conv2 = ir_1w1a.IRConv2d(planes,
planes,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.shortcut = nn.Sequential()
if stride != 1 or in_planes != planes:
if option == 'A':
"""
For CIFAR10 ResNet paper uses option A.
"""
self.shortcut = LambdaLayer(lambda x: F.pad(
x[:, :, ::2, ::2],
(0, 0, 0, 0, planes // 4, planes // 4), "constant", 0))
elif option == 'B':
self.shortcut = nn.Sequential(
ir_1w1a.IRConv2d(in_planes,
self.expansion * planes,
kernel_size=1,
stride=stride,
bias=False),
nn.BatchNorm2d(self.expansion * planes))
def __init__(self, num_classes=10):
super(VGG_SMALL_bi_1W1A, self).__init__()
self.ch_bi = change_to_bi()
#self.conv0 = ir_1w1a.my_Conv2d(3, 128, kernel_size=(
# 3, 24), padding=(1, 8), stride=(1, 8), bias=False)
self.conv0 = ir_1w1a.my_Conv2d(3,
128,
kernel_size=(1, 8),
stride=(1, 8),
bias=False)
# self.hold_to_bi = hold_to_bi(32, 32, inplanes=3, expansion=3)
# self.conv0 = ir_1w1a.IRConv2d(
# 9, 128, kernel_size=1, padding=0, bias=False)
# self.conv0 = nn.Conv2d(3, 128, kernel_size=1, padding=0, bias=False)
# self.conv0 = nn.Conv2d(3, 128, kernel_size=3, padding=1, bias=False)
# self.conv0 = ir_1w1a.IRConv2d(
# 3, 128, kernel_size = 1, padding = 1, bias = False)
self.bn0 = nn.BatchNorm2d(128)
self.conv1 = ir_1w1a.IRConv2d(128,
128,
kernel_size=3,
padding=1,
bias=False)
self.pooling = nn.MaxPool2d(kernel_size=2, stride=2)
self.bn1 = nn.BatchNorm2d(128)
# self.nonlinear = nn.ReLU(inplace=True)
self.nonlinear = nn.Hardtanh(inplace=True)
self.conv2 = ir_1w1a.IRConv2d(128,
256,
kernel_size=3,
padding=1,
bias=False)
self.bn2 = nn.BatchNorm2d(256)
self.conv3 = ir_1w1a.IRConv2d(256,
256,
kernel_size=3,
padding=1,
bias=False)
self.bn3 = nn.BatchNorm2d(256)
self.conv4 = ir_1w1a.IRConv2d(256,
512,
kernel_size=3,
padding=1,
bias=False)
self.bn4 = nn.BatchNorm2d(512)
self.conv5 = ir_1w1a.IRConv2d(512,
512,
kernel_size=3,
padding=1,
bias=False)
self.bn5 = nn.BatchNorm2d(512)
self.fc = nn.Linear(512 * 4 * 4, num_classes)
self._initialize_weights()
def __init__(self, num_classes=10):
super(VGG_SMALL_1W1A, self).__init__()
self.conv0 = nn.Conv2d(3, 128, kernel_size=3, padding=1, bias=False)
self.bn0 = nn.BatchNorm2d(128)
self.conv1 = ir_1w1a.IRConv2d(128, 128, kernel_size=3, padding=1, bias=False)
self.pooling = nn.MaxPool2d(kernel_size=2, stride=2)
self.bn1 = nn.BatchNorm2d(128)
# self.nonlinear = nn.ReLU(inplace=True)
self.nonlinear = nn.Hardtanh(inplace=True)
self.conv2 = ir_1w1a.IRConv2d(128, 256, kernel_size=3, padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(256)
self.conv3 = ir_1w1a.IRConv2d(256, 256, kernel_size=3, padding=1, bias=False)
self.bn3 = nn.BatchNorm2d(256)
self.conv4 = ir_1w1a.IRConv2d(256, 512, kernel_size=3, padding=1, bias=False)
self.bn4 = nn.BatchNorm2d(512)
self.conv5 = ir_1w1a.IRConv2d(512, 512, kernel_size=3, padding=1, bias=False)
self.bn5 = nn.BatchNorm2d(512)
self.fc = nn.Linear(512*4*4, num_classes)
self._initialize_weights()