forked from xxxnell/spatial-smoothing
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathresnet.py
203 lines (141 loc) · 8.27 KB
/
resnet.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
import torch
import torch.nn as nn
import models.layers as layers
import models.resnet_dnn_block as resnet_dnn
import models.resnet_mcdo_block as resnet_mcdo
import models.smoothing_block as smoothing
import models.classifier_block as classifier
class ResNet(nn.Module):
def __init__(self,
block, num_blocks,
sblock=smoothing.TanhBlurBlock, num_sblocks=(0, 0, 0, 0),
cblock=classifier.GAPBlock,
num_classes=10, stem=True, name="resnet", **block_kwargs):
super(ResNet, self).__init__()
self.name = name
self.layer0 = []
if stem:
self.layer0.append(layers.convnxn(3, 64, kernel_size=7, stride=2, padding=3))
self.layer0.append(layers.bn(64))
self.layer0.append(layers.relu())
self.layer0.append(nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
else:
self.layer0.append(layers.conv3x3(3, 64, stride=1))
self.layer0.append(layers.bn(64))
self.layer0.append(layers.relu())
self.layer0 = nn.Sequential(*self.layer0)
self.layer1 = self._make_layer(block, 64, 64, num_blocks[0], stride=1, **block_kwargs)
self.layer2 = self._make_layer(block, 64 * block.expansion, 128, num_blocks[1], stride=2, **block_kwargs)
self.layer3 = self._make_layer(block, 128 * block.expansion, 256, num_blocks[2], stride=2, **block_kwargs)
self.layer4 = self._make_layer(block, 256 * block.expansion, 512, num_blocks[3], stride=2, **block_kwargs)
self.smooth1 = self._make_smooth_layer(sblock, 64 * block.expansion, num_sblocks[0], **block_kwargs)
self.smooth2 = self._make_smooth_layer(sblock, 128 * block.expansion, num_sblocks[1], **block_kwargs)
self.smooth3 = self._make_smooth_layer(sblock, 256 * block.expansion, num_sblocks[2], **block_kwargs)
self.smooth4 = self._make_smooth_layer(sblock, 512 * block.expansion, num_sblocks[3], **block_kwargs)
self.classifier = []
if cblock is classifier.MLPBlock:
self.classifier.append(nn.AdaptiveAvgPool2d((7, 7)))
self.classifier.append(cblock(7 * 7 * 512 * block.expansion, num_classes, **block_kwargs))
else:
self.classifier.append(cblock(512 * block.expansion, num_classes, **block_kwargs))
self.classifier = nn.Sequential(*self.classifier)
@staticmethod
def _make_layer(block, in_channels, out_channels, num_blocks, stride, **block_kwargs):
stride_arr = [stride] + [1] * (num_blocks - 1)
layers, channels = [], in_channels
for stride in stride_arr:
layers.append(block(channels, out_channels, stride=stride, **block_kwargs))
channels = out_channels * block.expansion
return nn.Sequential(*layers)
@staticmethod
def _make_smooth_layer(sblock, in_filters, num_blocks, **block_kwargs):
layers = []
for _ in range(num_blocks):
layers.append(sblock(in_filters=in_filters, **block_kwargs))
return nn.Sequential(*layers)
def forward(self, x):
x = self.layer0(x)
x = self.layer1(x)
x = self.smooth1(x)
x = self.layer2(x)
x = self.smooth2(x)
x = self.layer3(x)
x = self.smooth3(x)
x = self.layer4(x)
x = self.smooth4(x)
x = self.classifier(x)
return x
# Deterministic
def dnn_18(num_classes=10, stem=True, name="resnet_dnn_18", **block_kwargs):
return ResNet(resnet_dnn.BasicBlock, [2, 2, 2, 2],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_34(num_classes=10, stem=True, name="resnet_dnn_34", **block_kwargs):
return ResNet(resnet_dnn.BasicBlock, [3, 4, 6, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_50(num_classes=10, stem=True, name="resnet_dnn_50", **block_kwargs):
return ResNet(resnet_dnn.Bottleneck, [3, 4, 6, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_101(num_classes=10, stem=True, name="resnet_dnn_101", **block_kwargs):
return ResNet(resnet_dnn.Bottleneck, [3, 4, 23, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_152(num_classes=10, stem=True, name="resnet_dnn_152", **block_kwargs):
return ResNet(resnet_dnn.Bottleneck, [3, 8, 36, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
# MC dropout
def mcdo_18(num_classes=10, stem=True, name="resnet_mcdo_18", **block_kwargs):
return ResNet(resnet_mcdo.BasicBlock, [2, 2, 2, 2],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_34(num_classes=10, stem=True, name="resnet_mcdo_34", **block_kwargs):
return ResNet(resnet_mcdo.BasicBlock, [3, 4, 6, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_50(num_classes=10, stem=True, name="resnet_mcdo_50", **block_kwargs):
return ResNet(resnet_mcdo.Bottleneck, [3, 4, 6, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_101(num_classes=10, stem=True, name="resnet_mcdo_101", **block_kwargs):
return ResNet(resnet_mcdo.Bottleneck, [3, 4, 23, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_152(num_classes=10, stem=True, name="resnet_mcdo_152", **block_kwargs):
return ResNet(resnet_mcdo.Bottleneck, [3, 8, 36, 3],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
# Deterministic + Smoothing
def dnn_smooth_18(num_classes=10, stem=True, name="resnet_dnn_smoothing_18", **block_kwargs):
return ResNet(resnet_dnn.BasicBlock, [2, 2, 2, 2],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_smooth_34(num_classes=10, stem=True, name="resnet_dnn_smoothing_34", **block_kwargs):
return ResNet(resnet_dnn.BasicBlock, [3, 4, 6, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_smooth_50(num_classes=10, stem=True, name="resnet_dnn_smoothing_50", **block_kwargs):
return ResNet(resnet_dnn.Bottleneck, [3, 4, 6, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_smooth_101(num_classes=10, stem=True, name="resnet_dnn_smoothing_101", **block_kwargs):
return ResNet(resnet_dnn.Bottleneck, [3, 4, 23, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def dnn_smooth_152(num_classes=10, stem=True, name="resnet_dnn_smoothing_152", **block_kwargs):
return ResNet(resnet_dnn.Bottleneck, [3, 8, 36, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
# MC dropout + Smoothing
def mcdo_smooth_18(num_classes=10, stem=True, name="resnet_mcdo_smoothing_18", **block_kwargs):
return ResNet(resnet_mcdo.BasicBlock, [2, 2, 2, 2],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_smooth_34(num_classes=10, stem=True, name="resnet_mcdo_smoothing_34", **block_kwargs):
return ResNet(resnet_mcdo.BasicBlock, [3, 4, 6, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_smooth_50(num_classes=10, stem=True, name="resnet_mcdo_smoothing_50", **block_kwargs):
return ResNet(resnet_mcdo.Bottleneck, [3, 4, 6, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_smooth_101(num_classes=10, stem=True, name="resnet_mcdo_smoothing_101", **block_kwargs):
return ResNet(resnet_mcdo.Bottleneck, [3, 4, 23, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)
def mcdo_smooth_152(num_classes=10, stem=True, name="resnet_mcdo_smoothing_152", **block_kwargs):
return ResNet(resnet_mcdo.Bottleneck, [3, 8, 36, 3],
num_sblocks=[1, 1, 1, 1],
num_classes=num_classes, stem=stem, name=name, **block_kwargs)