-
Notifications
You must be signed in to change notification settings - Fork 38
/
test_group_renorm.py
249 lines (198 loc) · 7.76 KB
/
test_group_renorm.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
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
import pytest
import numpy as np
from numpy.testing import assert_allclose
from keras.layers import Input
from keras import regularizers
from keras.utils.test_utils import layer_test, keras_test
from keras.models import Sequential, Model
from keras import backend as K
from group_norm import GroupNormalization
input_1 = np.arange(10)
input_2 = np.zeros(10)
input_3 = np.ones((10))
input_shapes = [np.ones((10, 10)), np.ones((10, 10, 10))]
@keras_test
def test_basic_groupnorm():
layer_test(GroupNormalization,
kwargs={'groups': 2,
'epsilon': 0.1,
'gamma_regularizer': regularizers.l2(0.01),
'beta_regularizer': regularizers.l2(0.01)},
input_shape=(3, 4, 2))
layer_test(GroupNormalization,
kwargs={'groups': 2,
'epsilon': 0.1,
'axis': 1},
input_shape=(3, 4, 2))
layer_test(GroupNormalization,
kwargs={'groups': 2,
'gamma_initializer': 'ones',
'beta_initializer': 'ones'},
input_shape=(3, 4, 2, 4))
if K.backend() != 'theano':
layer_test(GroupNormalization,
kwargs={'groups': 2,
'axis': 1,
'scale': False,
'center': False},
input_shape=(3, 4, 2, 4))
@keras_test
def test_groupnorm_correctness_1d():
model = Sequential()
norm = GroupNormalization(input_shape=(10,), groups=2)
model.add(norm)
model.compile(loss='mse', optimizer='rmsprop')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10))
model.fit(x, x, epochs=5, verbose=0)
out = model.predict(x)
out -= K.eval(norm.beta)
out /= K.eval(norm.gamma)
assert_allclose(out.mean(), 0.0, atol=1e-1)
assert_allclose(out.std(), 1.0, atol=1e-1)
@keras_test
def test_groupnorm_correctness_2d():
model = Sequential()
norm = GroupNormalization(axis=1, input_shape=(10, 6), groups=2)
model.add(norm)
model.compile(loss='mse', optimizer='rmsprop')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10, 6))
model.fit(x, x, epochs=5, verbose=0)
out = model.predict(x)
out -= np.reshape(K.eval(norm.beta), (1, 10, 1))
out /= np.reshape(K.eval(norm.gamma), (1, 10, 1))
assert_allclose(out.mean(axis=(0, 2)), 0.0, atol=1.1e-1)
assert_allclose(out.std(axis=(0, 2)), 1.0, atol=1.1e-1)
@keras_test
def test_groupnorm_correctness_2d_different_groups():
norm1 = GroupNormalization(axis=1, input_shape=(10, 6), groups=2)
norm2 = GroupNormalization(axis=1, input_shape=(10, 6), groups=1)
norm3 = GroupNormalization(axis=1, input_shape=(10, 6), groups=10)
model = Sequential()
model.add(norm1)
model.compile(loss='mse', optimizer='rmsprop')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10, 6))
model.fit(x, x, epochs=5, verbose=0)
out = model.predict(x)
out -= np.reshape(K.eval(norm1.beta), (1, 10, 1))
out /= np.reshape(K.eval(norm1.gamma), (1, 10, 1))
assert_allclose(out.mean(axis=(0, 2)), 0.0, atol=1.1e-1)
assert_allclose(out.std(axis=(0, 2)), 1.0, atol=1.1e-1)
model = Sequential()
model.add(norm2)
model.compile(loss='mse', optimizer='rmsprop')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10, 6))
model.fit(x, x, epochs=5, verbose=0)
out = model.predict(x)
out -= np.reshape(K.eval(norm2.beta), (1, 10, 1))
out /= np.reshape(K.eval(norm2.gamma), (1, 10, 1))
assert_allclose(out.mean(axis=(0, 2)), 0.0, atol=1.1e-1)
assert_allclose(out.std(axis=(0, 2)), 1.0, atol=1.1e-1)
model = Sequential()
model.add(norm3)
model.compile(loss='mse', optimizer='rmsprop')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10, 6))
model.fit(x, x, epochs=5, verbose=0)
out = model.predict(x)
out -= np.reshape(K.eval(norm3.beta), (1, 10, 1))
out /= np.reshape(K.eval(norm3.gamma), (1, 10, 1))
assert_allclose(out.mean(axis=(0, 2)), 0.0, atol=1.1e-1)
assert_allclose(out.std(axis=(0, 2)), 1.0, atol=1.1e-1)
@keras_test
def test_groupnorm_mode_twice():
# This is a regression test for issue #4881 with the old
# batch normalization functions in the Theano backend.
model = Sequential()
model.add(GroupNormalization(input_shape=(10, 5, 5), axis=1, groups=2))
model.add(GroupNormalization(input_shape=(10, 5, 5), axis=1, groups=2))
model.compile(loss='mse', optimizer='sgd')
x = np.random.normal(loc=5.0, scale=10.0, size=(20, 10, 5, 5))
model.fit(x, x, epochs=1, verbose=0)
model.predict(x)
@keras_test
def test_groupnorm_convnet():
model = Sequential()
norm = GroupNormalization(axis=1, input_shape=(3, 4, 4), groups=3)
model.add(norm)
model.compile(loss='mse', optimizer='sgd')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 3, 4, 4))
model.fit(x, x, epochs=4, verbose=0)
out = model.predict(x)
out -= np.reshape(K.eval(norm.beta), (1, 3, 1, 1))
out /= np.reshape(K.eval(norm.gamma), (1, 3, 1, 1))
assert_allclose(np.mean(out, axis=(0, 2, 3)), 0.0, atol=1e-1)
assert_allclose(np.std(out, axis=(0, 2, 3)), 1.0, atol=1e-1)
@keras_test
@pytest.mark.skipif((K.backend() == 'theano'),
reason='Bug with theano backend')
def test_groupnorm_convnet_no_center_no_scale():
model = Sequential()
norm = GroupNormalization(axis=-1, center=False, scale=False,
input_shape=(3, 4, 4), groups=2)
model.add(norm)
model.compile(loss='mse', optimizer='sgd')
# centered on 5.0, variance 10.0
x = np.random.normal(loc=5.0, scale=10.0, size=(1000, 3, 4, 4))
model.fit(x, x, epochs=4, verbose=0)
out = model.predict(x)
assert_allclose(np.mean(out, axis=(0, 2, 3)), 0.0, atol=1e-1)
assert_allclose(np.std(out, axis=(0, 2, 3)), 1.0, atol=1e-1)
@keras_test
def test_shared_groupnorm():
'''Test that a GN layer can be shared
across different data streams.
'''
# Test single layer reuse
bn = GroupNormalization(input_shape=(10,), groups=2)
x1 = Input(shape=(10,))
bn(x1)
x2 = Input(shape=(10,))
y2 = bn(x2)
x = np.random.normal(loc=5.0, scale=10.0, size=(2, 10))
model = Model(x2, y2)
assert len(model.updates) == 0
model.compile('sgd', 'mse')
model.train_on_batch(x, x)
# Test model-level reuse
x3 = Input(shape=(10,))
y3 = model(x3)
new_model = Model(x3, y3)
assert len(model.updates) == 0
new_model.compile('sgd', 'mse')
new_model.train_on_batch(x, x)
@keras_test
def test_that_trainable_disables_updates():
val_a = np.random.random((10, 4))
val_out = np.random.random((10, 4))
a = Input(shape=(4,))
layer = GroupNormalization(input_shape=(4,), groups=2)
b = layer(a)
model = Model(a, b)
model.trainable = False
assert len(model.updates) == 0
model.compile('sgd', 'mse')
assert len(model.updates) == 0
x1 = model.predict(val_a)
model.train_on_batch(val_a, val_out)
x2 = model.predict(val_a)
assert_allclose(x1, x2, atol=1e-7)
model.trainable = True
model.compile('sgd', 'mse')
assert len(model.updates) == 0
model.train_on_batch(val_a, val_out)
x2 = model.predict(val_a)
assert np.abs(np.sum(x1 - x2)) > 1e-5
layer.trainable = False
model.compile('sgd', 'mse')
assert len(model.updates) == 0
x1 = model.predict(val_a)
model.train_on_batch(val_a, val_out)
x2 = model.predict(val_a)
assert_allclose(x1, x2, atol=1e-7)
if __name__ == '__main__':
pytest.main([__file__])