-
-
Notifications
You must be signed in to change notification settings - Fork 2
/
mobilenet_v2_tsm.py
196 lines (168 loc) · 6.75 KB
/
mobilenet_v2_tsm.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
import torch.nn as nn
import torch
import math
def conv_bn(inp, oup, stride):
return nn.Sequential(
nn.Conv2d(inp, oup, 3, stride, 1, bias=False),
nn.BatchNorm2d(oup),
nn.ReLU6(inplace=True)
)
def conv_1x1_bn(inp, oup):
return nn.Sequential(
nn.Conv2d(inp, oup, 1, 1, 0, bias=False),
nn.BatchNorm2d(oup),
nn.ReLU6(inplace=True)
)
def make_divisible(x, divisible_by=8):
import numpy as np
return int(np.ceil(x * 1. / divisible_by) * divisible_by)
class InvertedResidual(nn.Module):
def __init__(self, inp, oup, stride, expand_ratio):
super(InvertedResidual, self).__init__()
self.stride = stride
assert stride in [1, 2]
hidden_dim = int(inp * expand_ratio)
self.use_res_connect = self.stride == 1 and inp == oup
if expand_ratio == 1:
self.conv = nn.Sequential(
# dw
nn.Conv2d(hidden_dim, hidden_dim, 3, stride, 1, groups=hidden_dim, bias=False),
nn.BatchNorm2d(hidden_dim),
nn.ReLU6(inplace=True),
# pw-linear
nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False),
nn.BatchNorm2d(oup),
)
else:
self.conv = nn.Sequential(
# pw
nn.Conv2d(inp, hidden_dim, 1, 1, 0, bias=False),
nn.BatchNorm2d(hidden_dim),
nn.ReLU6(inplace=True),
# dw
nn.Conv2d(hidden_dim, hidden_dim, 3, stride, 1, groups=hidden_dim, bias=False),
nn.BatchNorm2d(hidden_dim),
nn.ReLU6(inplace=True),
# pw-linear
nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False),
nn.BatchNorm2d(oup),
)
def forward(self, x):
if self.use_res_connect:
return x + self.conv(x)
else:
return self.conv(x)
class InvertedResidualWithShift(nn.Module):
def __init__(self, inp, oup, stride, expand_ratio):
super(InvertedResidualWithShift, self).__init__()
self.stride = stride
assert stride in [1, 2]
assert expand_ratio > 1
hidden_dim = int(inp * expand_ratio)
self.use_res_connect = self.stride == 1 and inp == oup
assert self.use_res_connect
self.conv = nn.Sequential(
# pw
nn.Conv2d(inp, hidden_dim, 1, 1, 0, bias=False),
nn.BatchNorm2d(hidden_dim),
nn.ReLU6(inplace=True),
# dw
nn.Conv2d(hidden_dim, hidden_dim, 3, stride, 1, groups=hidden_dim, bias=False),
nn.BatchNorm2d(hidden_dim),
nn.ReLU6(inplace=True),
# pw-linear
nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False),
nn.BatchNorm2d(oup),
)
def forward(self, x, shift_buffer):
c = x.size(1)
x1, x2 = x[:, : c // 8], x[:, c // 8:]
return x + self.conv(torch.cat((shift_buffer, x2), dim=1)), x1
class MobileNetV2(nn.Module):
def __init__(self, n_class=1000, input_size=224, width_mult=1.):
super(MobileNetV2, self).__init__()
input_channel = 32
last_channel = 1280
interverted_residual_setting = [
# t, c, n, s
[1, 16, 1, 1],
[6, 24, 2, 2],
[6, 32, 3, 2],
[6, 64, 4, 2],
[6, 96, 3, 1],
[6, 160, 3, 2],
[6, 320, 1, 1],
]
# building first layer
assert input_size % 32 == 0
# input_channel = make_divisible(input_channel * width_mult) # first channel is always 32!
self.last_channel = make_divisible(last_channel * width_mult) if width_mult > 1.0 else last_channel
self.features = [conv_bn(3, input_channel, 2)]
# building inverted residual blocks
global_idx = 0
shift_block_idx = [2, 4, 5, 7, 8, 9, 11, 12, 14, 15]
for t, c, n, s in interverted_residual_setting:
output_channel = make_divisible(c * width_mult) if t > 1 else c
for i in range(n):
if i == 0:
block = InvertedResidualWithShift if global_idx in shift_block_idx else InvertedResidual
self.features.append(block(input_channel, output_channel, s, expand_ratio=t))
global_idx += 1
else:
block = InvertedResidualWithShift if global_idx in shift_block_idx else InvertedResidual
self.features.append(block(input_channel, output_channel, 1, expand_ratio=t))
global_idx += 1
input_channel = output_channel
# building last several layers
self.features.append(conv_1x1_bn(input_channel, self.last_channel))
# make it nn.Sequential
self.features = nn.ModuleList(self.features)
# building classifier
self.classifier = nn.Linear(self.last_channel, n_class)
self._initialize_weights()
def forward(self, x, *shift_buffer):
shift_buffer_idx = 0
out_buffer = []
for f in self.features:
if isinstance(f, InvertedResidualWithShift):
x, s = f(x, shift_buffer[shift_buffer_idx])
shift_buffer_idx += 1
out_buffer.append(s)
else:
x = f(x)
x = x.mean(3).mean(2)
x = self.classifier(x)
return (x, *out_buffer)
def _initialize_weights(self):
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))
if m.bias is not None:
m.bias.data.zero_()
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.Linear):
n = m.weight.size(1)
m.weight.data.normal_(0, 0.01)
m.bias.data.zero_()
def mobilenet_v2_140():
return MobileNetV2(width_mult=1.4)
if __name__ == '__main__':
net = MobileNetV2()
x = torch.rand(1, 3, 224, 224)
shift_buffer = [torch.zeros([1, 3, 56, 56]),
torch.zeros([1, 4, 28, 28]),
torch.zeros([1, 4, 28, 28]),
torch.zeros([1, 8, 14, 14]),
torch.zeros([1, 8, 14, 14]),
torch.zeros([1, 8, 14, 14]),
torch.zeros([1, 12, 14, 14]),
torch.zeros([1, 12, 14, 14]),
torch.zeros([1, 20, 7, 7]),
torch.zeros([1, 20, 7, 7])]
with torch.no_grad():
for _ in range(10):
y, shift_buffer = net(x, *shift_buffer)
print([s.shape for s in shift_buffer])