forked from torchgan/torchgan
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest-models.py
179 lines (169 loc) · 7.39 KB
/
test-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
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
import unittest
import torch
import torch.distributions as distributions
import os
import sys
sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
from torchgan.models import *
class TestModels(unittest.TestCase):
def test_dcgan_generator(self):
encodings = [50, 100]
channels = [3, 4]
out_size = [32, 256]
step = [64, 128]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ReLU()]
last_nonlinearity = [None, torch.nn.LeakyReLU()]
for i in range(2):
gen = DCGANGenerator(encodings[i], out_size[i], channels[i], step[i],
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
z = torch.rand(10, encodings[i])
x = gen(z)
assert x.shape == (10, channels[i], out_size[i], out_size[i])
def test_dcgan_discriminator(self):
channels = [3, 4]
step = [64, 128]
in_size = [32, 64]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ReLU()]
last_nonlinearity = [None, torch.nn.LeakyReLU()]
for i in range(2):
dis = DCGANDiscriminator(in_size[i], channels[i], step[i],
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
x = torch.rand(10, channels[i], in_size[i], in_size[i])
loss = dis(x)
assert loss.shape == (10,)
def test_conditional_gan_generator(self):
encodings = [50, 100]
channels = [3, 4]
out_size = [32, 64]
classes = [5, 10]
step = [64, 128]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ELU(0.5)]
last_nonlinearity = [None, torch.nn.RReLU(0.25)]
for i in range(2):
ch = step[i]
x = torch.randn(10, encodings[i])
gen = ConditionalGANGenerator(classes[i], encodings[i], out_size[i], channels[i], ch,
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
labels = torch.randint(0, classes[i], (10,))
y = gen(x, labels)
assert y.shape == (10, channels[i], out_size[i], out_size[i])
def test_conditional_gan_discriminator(self):
channels = [3, 4]
in_size = [32, 64]
classes = [5, 10]
step = [64, 128]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ELU(0.5)]
last_nonlinearity = [None, torch.nn.RReLU(0.25)]
for i in range(2):
ch = step[i]
x = torch.randn(10, channels[i], in_size[i], in_size[i])
gen = ConditionalGANDiscriminator(classes[i], in_size[i], channels[i], ch,
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
labels = torch.randint(0, classes[i], (10,))
y = gen(x, labels)
assert y.shape == (10,)
def test_acgan_generator(self):
encodings = [50, 100]
channels = [3, 4]
out_size = [32, 64]
classes = [5, 10]
step = [64, 128]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ELU(0.5)]
last_nonlinearity = [None, torch.nn.RReLU(0.25)]
for i in range(2):
ch = step[i]
x = torch.randn(10, encodings[i])
gen = ACGANGenerator(classes[i], encodings[i], out_size[i], channels[i], ch,
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
labels = torch.randint(0, classes[i], (10,))
y = gen(x, labels)
assert y.shape == (10, channels[i], out_size[i], out_size[i])
def test_acgan_discriminator(self):
channels = [3, 4]
in_size = [32, 64]
classes = [5, 10]
step = [64, 128]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ELU(0.5)]
last_nonlinearity = [None, torch.nn.RReLU(0.25)]
for i in range(2):
ch = step[i]
x = torch.randn(10, channels[i], in_size[i], in_size[i])
gen = ACGANDiscriminator(classes[i], in_size[i], channels[i], ch,
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
dx, cx = gen(x, mode='combine')
assert dx.shape == (10,)
assert cx.shape == (10, classes[i])
def test_infogan_generator(self):
encodings = [50, 100]
channels = [3, 4]
out_size = [32, 64]
step = [64, 128]
dim_cont = [10, 20]
dim_dis = [30, 40]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ELU(0.5)]
last_nonlinearity = [None, torch.nn.RReLU(0.25)]
for i in range(2):
ch = step[i]
x = torch.randn(10, encodings[i])
cont = torch.rand(10, dim_cont[i])
dis = torch.zeros(10, dim_dis[i])
gen = InfoGANGenerator(dim_cont[i], dim_dis[i], encodings[i], out_size[i], channels[i],
ch, batchnorm[i], nonlinearities[i], last_nonlinearity[i])
y = gen(x, cont, dis)
assert y.shape == (10, channels[i], out_size[i], out_size[i])
def test_infogan_discriminator(self):
channels = [3, 4]
in_size = [32, 64]
dim_cont = [10, 20]
dim_dis = [30, 40]
step = [64, 128]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ELU(0.5)]
last_nonlinearity = [None, torch.nn.RReLU(0.25)]
for i in range(2):
x = torch.randn(10, channels[i], in_size[i], in_size[i])
D = InfoGANDiscriminator(dim_dis[i], dim_cont[i], in_size[i], channels[i], step[i],
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
y, dist_dis, dist_cont = D(x, True)
assert y.shape == (10, 1, 1, 1)
assert isinstance(dist_dis, distributions.OneHotCategorical)
assert isinstance(dist_cont, distributions.Normal)
assert dist_dis.sample().shape == (10, dim_dis[i])
assert dist_cont.sample().shape == (10, dim_cont[i])
def test_autoencoding_generator(self):
encodings = [50, 100]
channels = [3, 4]
out_size = [32, 64]
step = [64, 128]
scale = [2, 2]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ReLU()]
last_nonlinearity = [None, torch.nn.LeakyReLU()]
for i in range(2):
gen = AutoEncodingGenerator(encodings[i], out_size[i], channels[i], step[i], scale[i],
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
z = torch.rand(10, encodings[i])
x = gen(z)
assert x.shape == (10, channels[i], out_size[i], out_size[i])
def test_autoencoding_discriminator(self):
channels = [3, 4]
encodings = [50, 100]
in_size = [32, 64]
step = [64, 128]
scale = [2, 2]
batchnorm = [True, False]
nonlinearities = [None, torch.nn.ReLU()]
last_nonlinearity = [None, torch.nn.LeakyReLU()]
for i in range(2):
dis = AutoEncodingDiscriminator(in_size[i], channels[i], encodings[i], step[i], scale[i],
batchnorm[i], nonlinearities[i], last_nonlinearity[i])
x = torch.rand(10, channels[i], in_size[i], in_size[i])
loss = dis(x)
assert loss.shape == (10,)