def conv1x1(in_planes, out_planes, stride=1):
"""1x1 convolution"""
return adder.adder2d(in_planes,
out_planes,
kernel_size=1,
stride=stride,
bias=False)
def conv3x3(in_planes, out_planes, stride=1):
" 3x3 convolution with padding "
return adder.adder2d(in_planes,
out_planes,
kernel_size=3,
stride=stride,
padding=1,
bias=False)
def _make_layer(self, block, planes, blocks, stride=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
adder.adder2d(self.inplanes, planes * block.expansion, kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(planes * block.expansion)
)
layers = [block(inplanes=self.inplanes, planes=planes, stride=stride, downsample=downsample)]
self.inplanes = planes * block.expansion
for _ in range(1, blocks):
layers.append(block(inplanes=self.inplanes, planes=planes))
return nn.Sequential(*layers)
def conv3x3(input_channel, output_channel, stride=1, op=1):
" 3x3 convolution with padding op:1-add 0:mul"
if op:
return adder.adder2d(input_channel,
output_channel,
kernel_size=3,
stride=stride,
padding=1,
bias=False)
else:
return nn.Conv2d(input_channel,
output_channel,
kernel_size=3,
stride=stride,
padding=1,
bias=False)
def conv3x3(in_channels, out_channels, padding=1):
"""3x3 convolution with padding"""
return adder.adder2d(in_channels, out_channels, kernel_size=3, stride=1,
padding=1, bias=False)
