in models/cifar/resnext_cnsn.py [0:0]
def __init__(self, depth, cardinality, base_width, num_classes,
active_num=None, pos=None, beta=None,
crop=None, cnsn_type=None):
super(CifarResNeXt, self).__init__()
norm_func = nn.BatchNorm2d
if beta is not None:
print('beta: {}'.format(beta))
if crop is not None:
print('crop mode: {}'.format(crop))
# Model type specifies number of layers for CIFAR-10 and CIFAR-100 model
assert (depth - 2) % 9 == 0, 'depth should be one of 29, 38, 47, 56, 101'
layer_blocks = (depth - 2) // 9
self.cardinality = cardinality
self.base_width = base_width
self.num_classes = num_classes
self.conv_1_3x3 = nn.Conv2d(3, 64, 3, 1, 1, bias=False)
self.bn_1 = norm_func(64)
self.inplanes = 64
# 1st block
self.stage_1 = self._make_layer_cnsn(ResNeXtBottleneckCustom, 64, layer_blocks, norm_func,
pos=pos, beta=beta, crop=crop,
cnsn_type=cnsn_type, stride=1)
# 2nd block
self.stage_2 = self._make_layer_cnsn(ResNeXtBottleneckCustom, 128, layer_blocks, norm_func,
pos=pos, beta=beta, crop=crop,
cnsn_type=cnsn_type, stride=2)
# 3rd block
self.stage_3 = self._make_layer_cnsn(ResNeXtBottleneckCustom, 256, layer_blocks, norm_func,
pos=pos, beta=beta, crop=crop,
cnsn_type=cnsn_type, stride=2)
self.avgpool = nn.AvgPool2d(8)
self.classifier = nn.Linear(256 * ResNeXtBottleneckCustom.expansion, num_classes)
self.cn_modules = []
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.Linear):
init.kaiming_normal_(m.weight)
m.bias.data.zero_()
elif isinstance(m, CrossNorm):
self.cn_modules.append(m)
if 'cn' in cnsn_type:
self.cn_num = len(self.cn_modules)
assert self.cn_num > 0
print('cn_num: {}'.format(self.cn_num))
self.active_num = active_num
assert self.active_num > 0
print('active_num: {}'.format(self.active_num))