MCPcopy Create free account
hub / github.com/thygate/stable-diffusion-webui-depthmap-script / __init__

Method __init__

lib/Resnext_torch.py:121–170  ·  view source on GitHub ↗
(self, block, layers, num_classes=1000, zero_init_residual=False,
                 groups=1, width_per_group=64, replace_stride_with_dilation=None,
                 norm_layer=None)

Source from the content-addressed store, hash-verified

119class ResNet(nn.Module):
120
121 def __init__(self, block, layers, num_classes=1000, zero_init_residual=False,
122 groups=1, width_per_group=64, replace_stride_with_dilation=None,
123 norm_layer=None):
124 super(ResNet, self).__init__()
125 if norm_layer is None:
126 norm_layer = nn.BatchNorm2d
127 self._norm_layer = norm_layer
128
129 self.inplanes = 64
130 self.dilation = 1
131 if replace_stride_with_dilation is None:
132 # each element in the tuple indicates if we should replace
133 # the 2x2 stride with a dilated convolution instead
134 replace_stride_with_dilation = [False, False, False]
135 if len(replace_stride_with_dilation) != 3:
136 raise ValueError("replace_stride_with_dilation should be None "
137 "or a 3-element tuple, got {}".format(replace_stride_with_dilation))
138 self.groups = groups
139 self.base_width = width_per_group
140 self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3,
141 bias=False)
142 self.bn1 = norm_layer(self.inplanes)
143 self.relu = nn.ReLU(inplace=True)
144 self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
145 self.layer1 = self._make_layer(block, 64, layers[0])
146 self.layer2 = self._make_layer(block, 128, layers[1], stride=2,
147 dilate=replace_stride_with_dilation[0])
148 self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
149 dilate=replace_stride_with_dilation[1])
150 self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
151 dilate=replace_stride_with_dilation[2])
152 #self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
153 #self.fc = nn.Linear(512 * block.expansion, num_classes)
154
155 for m in self.modules():
156 if isinstance(m, nn.Conv2d):
157 nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
158 elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
159 nn.init.constant_(m.weight, 1)
160 nn.init.constant_(m.bias, 0)
161
162 # Zero-initialize the last BN in each residual branch,
163 # so that the residual branch starts with zeros, and each residual block behaves like an identity.
164 # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
165 if zero_init_residual:
166 for m in self.modules():
167 if isinstance(m, Bottleneck):
168 nn.init.constant_(m.bn3.weight, 0)
169 elif isinstance(m, BasicBlock):
170 nn.init.constant_(m.bn2.weight, 0)
171
172 def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
173 norm_layer = self._norm_layer

Callers 2

__init__Method · 0.45
__init__Method · 0.45

Calls 2

_make_layerMethod · 0.95
norm_layerFunction · 0.85

Tested by

no test coverage detected