-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathresnet.py
189 lines (143 loc) · 8.8 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
import torch
import torch.nn as nn
import torchvision
import torch.nn.utils.prune as prune
pretrained_resnet = torchvision.models.resnet18(pretrained = True)
class BasicBlock(nn.Module):
"""Basic Block for resnet
"""
expansion = 1
def __init__(self, in_channels, out_channels, stride=1):
super().__init__()
#residual function
self.residual_function = nn.Sequential(
nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=False),
nn.BatchNorm2d(out_channels),
nn.ReLU(inplace=True),
nn.Conv2d(out_channels, out_channels * BasicBlock.expansion, kernel_size=3, padding=1, bias=False),
nn.BatchNorm2d(out_channels * BasicBlock.expansion)
)
#shortcut
self.shortcut = nn.Sequential()
if stride != 1 or in_channels != BasicBlock.expansion * out_channels:
self.shortcut = nn.Sequential(
nn.Conv2d(in_channels, out_channels * BasicBlock.expansion, kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(out_channels * BasicBlock.expansion)
)
def forward(self, x):
return nn.ReLU(inplace=True)(self.residual_function(x) + self.shortcut(x))
class BottleNeck(nn.Module):
"""Residual block for resnet over 50 layers
"""
expansion = 4
def __init__(self, in_channels, out_channels, stride=1):
super().__init__()
self.residual_function = nn.Sequential(
nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False),
nn.BatchNorm2d(out_channels),
nn.ReLU(inplace=True),
nn.Conv2d(out_channels, out_channels, stride=stride, kernel_size=3, padding=1, bias=False),
nn.BatchNorm2d(out_channels),
nn.ReLU(inplace=True),
nn.Conv2d(out_channels, out_channels * BottleNeck.expansion, kernel_size=1, bias=False),
nn.BatchNorm2d(out_channels * BottleNeck.expansion),
)
self.shortcut = nn.Sequential()
if stride != 1 or in_channels != out_channels * BottleNeck.expansion:
self.shortcut = nn.Sequential(
nn.Conv2d(in_channels, out_channels * BottleNeck.expansion, stride=stride, kernel_size=1, bias=False),
nn.BatchNorm2d(out_channels * BottleNeck.expansion)
)
def forward(self, x):
return nn.ReLU(inplace=True)(self.residual_function(x) + self.shortcut(x))
class ResNet(nn.Module):
def __init__(self, block, num_block=4, num_classes=10):
super().__init__()
self.in_channels = 64
self.conv1 = nn.Sequential(
nn.Conv2d(3, 64, kernel_size=3, padding=1, bias=False),
nn.BatchNorm2d(64),
nn.ReLU(inplace=True))
self.conv2_x = pretrained_resnet.layer1 #self._make_layer(block, 64, num_block[0], 1)
self.conv3_x = pretrained_resnet.layer2 #self._make_layer(block, 128, num_block[1], 2)
self.conv4_x = pretrained_resnet.layer3 #self._make_layer(block, 256, num_block[2], 2)
self.conv5_x = pretrained_resnet.layer4 #self._make_layer(block, 512, num_block[3], 2)
self.avg_pool = nn.AdaptiveAvgPool2d((1, 1))
self.fc = nn.Linear(512 * block.expansion, num_classes)
def forward(self, x):
output = self.conv1(x)
output = self.conv2_x(output)
output = self.conv3_x(output)
output = self.conv4_x(output)
output = self.conv5_x(output)
output = self.avg_pool(output)
output = output.view(output.size(0), -1)
output = self.fc(output)
return output
def _prune_res_unit1(self,ratio = 0.1):
prune.ln_structured(list(self.conv1)[0], name="weight", amount=ratio, n=1, dim=0)
def _prune_res_unit2(self,ratio = 0.1):
prune.ln_structured(list(self.conv2_x)[0].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv2_x)[0].conv2, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv2_x)[1].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv2_x)[1].conv2, name="weight", amount=ratio, n=1, dim=0)
def _prune_res_unit3(self,ratio = 0.1):
prune.ln_structured(list(self.conv3_x)[0].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv3_x)[0].conv2, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv3_x)[1].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv3_x)[1].conv2, name="weight", amount=ratio, n=1, dim=0)
def _prune_res_unit4(self,ratio = 0.1):
prune.ln_structured(list(self.conv4_x)[0].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv4_x)[0].conv2, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv4_x)[1].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv4_x)[1].conv2, name="weight", amount=ratio, n=1, dim=0)
def _prune_res_unit5(self,ratio = 0.1):
prune.ln_structured(list(self.conv5_x)[0].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv5_x)[0].conv2, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv5_x)[1].conv1, name="weight", amount=ratio, n=1, dim=0)
prune.ln_structured(list(self.conv5_x)[1].conv2, name="weight", amount=ratio, n=1, dim=0)
def _remove_res_unit1(self):
list(self.conv1)[0].weight_mask.data = list(self.conv1)[0].weight_mask.data * 0 + 1
prune.remove(list(self.conv1)[0], name="weight")
def _remove_res_unit2(self):
list(self.conv2_x)[0].conv1.weight_mask.data = list(self.conv2_x)[0].conv1.weight_mask.data * 0 + 1
list(self.conv2_x)[0].conv2.weight_mask.data = list(self.conv2_x)[0].conv2.weight_mask.data * 0 + 1
list(self.conv2_x)[1].conv1.weight_mask.data = list(self.conv2_x)[1].conv1.weight_mask.data * 0 + 1
list(self.conv2_x)[1].conv2.weight_mask.data = list(self.conv2_x)[1].conv2.weight_mask.data * 0 + 1
prune.remove(list(self.conv2_x)[0].conv1, name="weight")
prune.remove(list(self.conv2_x)[0].conv2, name="weight")
prune.remove(list(self.conv2_x)[1].conv1, name="weight")
prune.remove(list(self.conv2_x)[1].conv2, name="weight")
def _remove_res_unit3(self):
list(self.conv3_x)[0].conv1.weight_mask.data = list(self.conv3_x)[0].conv1.weight_mask.data * 0 + 1
list(self.conv3_x)[0].conv2.weight_mask.data = list(self.conv3_x)[0].conv2.weight_mask.data * 0 + 1
list(self.conv3_x)[1].conv1.weight_mask.data = list(self.conv3_x)[1].conv1.weight_mask.data * 0 + 1
list(self.conv3_x)[1].conv2.weight_mask.data = list(self.conv3_x)[1].conv2.weight_mask.data * 0 + 1
prune.remove(list(self.conv3_x)[0].conv1, name="weight")
prune.remove(list(self.conv3_x)[0].conv2, name="weight")
prune.remove(list(self.conv3_x)[1].conv1, name="weight")
prune.remove(list(self.conv3_x)[1].conv2, name="weight")
def _remove_res_unit4(self):
list(self.conv4_x)[0].conv1.weight_mask.data = list(self.conv4_x)[0].conv1.weight_mask.data * 0 + 1
list(self.conv4_x)[0].conv2.weight_mask.data = list(self.conv4_x)[0].conv2.weight_mask.data * 0 + 1
list(self.conv4_x)[1].conv1.weight_mask.data = list(self.conv4_x)[1].conv1.weight_mask.data * 0 + 1
list(self.conv4_x)[1].conv2.weight_mask.data = list(self.conv4_x)[1].conv2.weight_mask.data * 0 + 1
prune.remove(list(self.conv4_x)[0].conv1, name="weight")
prune.remove(list(self.conv4_x)[0].conv2, name="weight")
prune.remove(list(self.conv4_x)[1].conv1, name="weight")
prune.remove(list(self.conv4_x)[1].conv2, name="weight")
def _remove_res_unit5(self):
list(self.conv5_x)[0].conv1.weight_mask.data = list(self.conv5_x)[0].conv1.weight_mask.data * 0 + 1
list(self.conv5_x)[0].conv2.weight_mask.data = list(self.conv5_x)[0].conv2.weight_mask.data * 0 + 1
list(self.conv5_x)[1].conv1.weight_mask.data = list(self.conv5_x)[1].conv1.weight_mask.data * 0 + 1
list(self.conv5_x)[1].conv2.weight_mask.data = list(self.conv5_x)[1].conv2.weight_mask.data * 0 + 1
prune.remove(list(self.conv5_x)[0].conv1, name="weight")
prune.remove(list(self.conv5_x)[0].conv2, name="weight")
prune.remove(list(self.conv5_x)[1].conv1, name="weight")
prune.remove(list(self.conv5_x)[1].conv2, name="weight")
def resnet18():
""" return a ResNet 18 object
"""
return ResNet(BasicBlock, [2, 2, 2, 2])
if __name__ == '__main__':
model = resnet18()