-
Notifications
You must be signed in to change notification settings - Fork 0
/
preactresnet.py
138 lines (109 loc) · 5.71 KB
/
preactresnet.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
'''Pre-activation ResNet in PyTorch.
Reference:
[1] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
Identity Mappings in Deep Residual Networks. arXiv:1603.05027
'''
import torch
import torch.nn as nn
import torch.nn.functional as F
from cayley_ortho_conv import Cayley
from block_ortho_conv import BCOP
from skew_ortho_conv import SOC
from custom_activations import *
from utils import conv_mapping, activation_mapping
class PreActBlock(nn.Module):
'''Pre-activation version of the BasicBlock.'''
expansion = 1
def __init__(self, in_planes, planes, conv_layer, activation_name, stride=1):
super(PreActBlock, self).__init__()
self.activation1 = activation_mapping(activation_name, channels=in_planes)
self.activation2 = activation_mapping(activation_name, channels=planes)
self.bn1 = nn.BatchNorm2d(in_planes)
self.conv1 = conv_layer(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.conv2 = conv_layer(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
if stride != 1 or in_planes != self.expansion*planes:
self.shortcut = nn.Sequential(
conv_layer(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False)
)
def forward(self, x):
out = self.activation1(self.bn1(x))
shortcut = self.shortcut(out) if hasattr(self, 'shortcut') else x
out = self.conv1(out)
out = self.conv2(self.activation2(self.bn2(out)))
out += shortcut
return out
class PreActBottleneck(nn.Module):
'''Pre-activation version of the original Bottleneck module.'''
expansion = 4
def __init__(self, in_planes, planes, conv_name, activation_name, stride=1):
super(PreActBottleneck, self).__init__()
self.activation1 = activation_mapping(activation_name, channels=in_planes)
self.activation2 = activation_mapping(activation_name, channels=planes)
self.activation3 = activation_mapping(activation_name, channels=planes)
self.bn1 = nn.BatchNorm2d(in_planes)
self.conv1 = conv_layer(in_planes, planes, kernel_size=1, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.conv2 = conv_layer(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
self.bn3 = nn.BatchNorm2d(planes)
self.conv3 = conv_layer(planes, self.expansion*planes, kernel_size=1, bias=False)
if stride != 1 or in_planes != self.expansion*planes:
self.shortcut = nn.Sequential(
conv_layer(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False)
)
def forward(self, x):
out = self.activation1(self.bn1(x))
shortcut = self.shortcut(out) if hasattr(self, 'shortcut') else x
out = self.conv1(out)
out = self.conv2(self.activation2(self.bn2(out)))
out = self.conv3(self.activation3(self.bn3(out)))
out += shortcut
return out
class PreActResNet(nn.Module):
def __init__(self, block, num_blocks, conv_name, activation_name, num_classes=10):
super(PreActResNet, self).__init__()
conv_layer = conv_mapping[conv_name]
self.in_planes = 64
self.conv1 = conv_layer(3, 64, kernel_size=3, stride=1, padding=1, bias=False)
self.layer1 = self._make_layer(block, 64, num_blocks[0], conv_layer, activation_name, stride=1)
self.layer2 = self._make_layer(block, 128, num_blocks[1], conv_layer, activation_name, stride=2)
self.layer3 = self._make_layer(block, 256, num_blocks[2], conv_layer, activation_name, stride=2)
self.layer4 = self._make_layer(block, 512, num_blocks[3], conv_layer, activation_name, stride=2)
self.bn = nn.BatchNorm2d(512 * block.expansion)
self.activation = activation_mapping(activation_name, channels=512 * block.expansion)
self.linear = nn.Linear(512 * block.expansion, num_classes)
def _make_layer(self, block, planes, num_blocks, conv_layer, activation_name, stride):
strides = [stride] + [1]*(num_blocks-1)
layers = []
for stride in strides:
layers.append(block(self.in_planes, planes, conv_layer, activation_name, stride))
self.in_planes = planes * block.expansion
return nn.Sequential(*layers)
def forward(self, x):
out = self.conv1(x)
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
out = self.layer4(out)
out = self.activation(self.bn(out))
out = F.avg_pool2d(out, 4)
out = out.view(out.size(0), -1)
out = self.linear(out)
return out
def PreActResNet18(conv_name='standard', activation_name='relu', num_classes=10):
return PreActResNet(PreActBlock, [2,2,2,2], conv_name, activation_name, num_classes)
def PreActResNet34(conv_name='standard', activation_name='relu', num_classes=10):
return PreActResNet(PreActBlock, [3,4,6,3], conv_name, activation_name, num_classes)
def PreActResNet50(conv_name='standard', activation_name='relu', num_classes=10):
return PreActResNet(PreActBottleneck, [3,4,6,3], conv_name, activation_name, num_classes)
def PreActResNet101(conv_name='standard', activation_name='relu', num_classes=10):
return PreActResNet(PreActBottleneck, [3,4,23,3], conv_name, activation_name, num_classes)
def PreActResNet152(conv_name='standard', activation_name='relu', num_classes=10):
return PreActResNet(PreActBottleneck, [3,8,36,3], conv_name, activation_name, num_classes)
resnet_mapping = {
'resnet18': PreActResNet18,
'resnet34': PreActResNet34,
'resnet50': PreActResNet50,
'resnet101': PreActResNet101,
'resnet152': PreActResNet152
}