def __init__(self, in_planes, planes, dropout_rate, stride=1, sample_ratio=1.0, minimal_k=1):
super(wide_basic, self).__init__()
self.bn1 = nn.BatchNorm2d(in_planes)
self.conv1 = approx_Conv2d_dist(in_planes, planes, kernel_size=3, padding=1, bias=True, sample_ratio=sample_ratio, minimal_k=minimal_k)
#self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, padding=1, bias=True)
self.dropout = nn.Dropout(p=dropout_rate)
self.bn2 = nn.BatchNorm2d(planes)
self.conv2 = approx_Conv2d_dist(planes, planes, kernel_size=3, stride=stride, padding=1, bias=True, sample_ratio=sample_ratio, minimal_k=minimal_k)
#self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=True)
self.shortcut = nn.Sequential()
if stride != 1 or in_planes != planes:
self.shortcut = nn.Sequential(
approx_Conv2d_dist(in_planes, planes, kernel_size=1, stride=stride, bias=True, sample_ratio=sample_ratio, minimal_k=minimal_k),
#nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride, bias=True),
)
def conv1x1(in_planes, out_planes, stride=1):
"""1x1 convolution"""
return approx_Conv2d_dist(in_planes,
out_planes,
kernel_size=1,
stride=stride,
bias=False,
sample_ratio=sample_ratio,
minimal_k=minimal_k)
def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1):
"""3x3 convolution with padding"""
return approx_Conv2d_dist(in_planes,
out_planes,
kernel_size=3,
stride=stride,
padding=dilation,
groups=groups,
bias=False,
dilation=dilation,
sample_ratio=sample_ratio,
minimal_k=minimal_k)
def conv3x3(in_planes, out_planes, stride=1, sample_ratio=1.0, minimal_k=1):
return approx_Conv2d_dist(in_planes, out_planes, kernel_size=3, stride=stride, padding=1, bias=True, sample_ratio=sample_ratio, minimal_k=minimal_k)