-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmodels.py
162 lines (137 loc) · 6.45 KB
/
models.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
import numpy as np
import torch.nn as nn
import torch.nn.functional as F
class DCGAN_G(nn.Module):
"""
DCGAN_G architecture, " Unsupervised Representation Learning with
Deep Convolutional Generative Adversarial Networks" (Radford et al.)
"""
def __init__(self, image_size, nz, image_depth, num_filters, num_extra_layers=0):
super(DCGAN_G, self).__init__()
assert image_size % 16 == 0, "isize has to be a multiple of 16"
cngf, tisize = num_filters//2, 4
while tisize != image_size:
cngf = cngf * 2
tisize = tisize * 2
main = nn.Sequential()
main.add_module('initial_{0}-{1}_convt'.format(nz, cngf),
nn.ConvTranspose2d(nz, cngf, 4, 1, 0, bias=False))
main.add_module('initial_{0}_batchnorm'.format(cngf),
nn.BatchNorm2d(cngf))
main.add_module('initial_{0}_relu'.format(cngf),
nn.ReLU(True))
csize = 4
while csize < image_size//2:
main.add_module('pyramid_{0}-{1}_convt'.format(cngf, cngf//2),
nn.ConvTranspose2d(cngf, cngf//2, 4, 2, 1, bias=False))
main.add_module('pyramid_{0}_batchnorm'.format(cngf//2),
nn.BatchNorm2d(cngf//2))
main.add_module('pyramid_{0}_relu'.format(cngf//2),
nn.ReLU(True))
cngf = cngf // 2
csize = csize * 2
# Extra layers
for t in range(num_extra_layers):
main.add_module('extra-layers-{0}_{1}_conv'.format(t, cngf),
nn.Conv2d(cngf, cngf, 3, 1, 1, bias=False))
main.add_module('extra-layers-{0}_{1}_batchnorm'.format(t, cngf),
nn.BatchNorm2d(cngf))
main.add_module('extra-layers-{0}_{1}_relu'.format(t, cngf),
nn.ReLU(True))
main.add_module('final_{0}-{1}_convt'.format(cngf, image_depth),
nn.ConvTranspose2d(cngf, image_depth, 4, 2, 1, bias=False))
main.add_module('final_{0}_tanh'.format(image_depth),
nn.Tanh())
self.main = main
def forward(self, x):
return self.main(x)
class DCGAN_D(nn.Module):
"""
DCGAN_D architecture, " Unsupervised Representation Learning with
Deep Convolutional Generative Adversarial Networks" (Radford et al.)
"""
def __init__(self, image_size, image_depth, num_filters, num_extra_layers=0):
super(DCGAN_D, self).__init__()
assert image_size % 16 == 0, "isize has to be a multiple of 16"
main = nn.Sequential()
# input is nc x isize x isize
main.add_module('initial_conv_{0}-{1}'.format(image_depth, num_filters),
nn.Conv2d(image_depth, num_filters, 4, 2, 1, bias=False))
main.add_module('initial_relu_{0}'.format(num_filters),
nn.LeakyReLU(0.2, inplace=True))
csize, cndf = image_size / 2, num_filters
# Extra layers
for t in range(num_extra_layers):
main.add_module('extra-layers-{0}_{1}_conv'.format(t, cndf),
nn.Conv2d(cndf, cndf, 3, 1, 1, bias=False))
main.add_module('extra-layers-{0}_{1}_batchnorm'.format(t, cndf),
nn.BatchNorm2d(cndf))
main.add_module('extra-layers-{0}_{1}_relu'.format(t, cndf),
nn.LeakyReLU(0.2, inplace=True))
while csize > 4:
in_feat = cndf
out_feat = cndf * 2
main.add_module('pyramid_{0}-{1}_conv'.format(in_feat, out_feat),
nn.Conv2d(in_feat, out_feat, 4, 2, 1, bias=False))
main.add_module('pyramid_{0}_batchnorm'.format(out_feat),
nn.BatchNorm2d(out_feat))
main.add_module('pyramid_{0}_relu'.format(out_feat),
nn.LeakyReLU(0.2, inplace=True))
cndf = cndf * 2
csize = csize / 2
# state size. K x 4 x 4
main.add_module('final_{0}-{1}_conv'.format(cndf, 1),
nn.Conv2d(cndf, 1, 4, 1, 0, bias=False))
self.main = main
def forward(self, x):
return self.main(x).view(-1,1)
class FC_leaky(nn.Module):
"""
Fully connected net with LeakyReLU activations
"""
def __init__(self, input_size=1, output_size=1, hidden_layer_size=512, a=0.5):
super(FC_leaky, self).__init__()
main = nn.Sequential(
nn.Linear(input_size, hidden_layer_size),
nn.LeakyReLU(a, inplace=True),
nn.Linear(hidden_layer_size, hidden_layer_size),
nn.LeakyReLU(a, inplace=True),
nn.Linear(hidden_layer_size, hidden_layer_size),
nn.LeakyReLU(a, inplace=True),
nn.Linear(hidden_layer_size, output_size),
)
self.main = main
for m in self.main.modules():
if isinstance(m, nn.Linear):
nn.init.kaiming_normal_(m.weight, a=a, mode='fan_in', nonlinearity='leaky_relu')
if m.bias is not None:
nn.init.zeros_(m.bias)
def forward(self, x):
return self.main(x)
class SELU(nn.Module):
def __init__(self):
super(SELU, self).__init__()
self.alpha = 1.6732632423543772848170429916717
self.scale = 1.0507009873554804934193349852946
def forward(self, x):
return self.scale * F.elu(x, self.alpha)
class FC_selu(nn.Module):
"""
Fully connected with selu activations
"""
def __init__(self, input_size=1, output_size=1, hidden_layer_size=512, num_extra_layers=2):
super(FC_selu, self).__init__()
main = nn.Sequential()
main.add_module('layer-0', nn.Sequential(nn.Linear(input_size, hidden_layer_size, bias=False), SELU()))
for i in range(num_extra_layers):
main.add_module('layer-{}'.format(i+1), nn.Sequential(nn.Linear(hidden_layer_size, hidden_layer_size, bias=False), SELU()))
main.add_module('layer-{}'.format(num_extra_layers+1), nn.Sequential(nn.Linear(hidden_layer_size, output_size, bias=True)))
for m in main.modules():
if isinstance(m, nn.Linear):
fan_in, _ = nn.init._calculate_fan_in_and_fan_out(m.weight)
nn.init.normal_(m.weight, 0, np.sqrt(1./fan_in))
if m.bias is not None:
nn.init.zeros_(m.bias)
self.main = main
def forward(self, x):
return self.main(x)