-
Notifications
You must be signed in to change notification settings - Fork 1
/
convnet.py
179 lines (154 loc) · 5.21 KB
/
convnet.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
"""Classifying MNIST images by a convolutional network"""
import gzip
import os
import numpy as np
import autograd as ag
random_state = np.random.RandomState(0)
def read_mnist_labels(fn):
with gzip.open(fn, 'rb') as f:
content = f.read()
num_images = int.from_bytes(content[4:8], byteorder='big')
labels = np.zeros((num_images, 10), dtype=np.float32)
indices = np.frombuffer(content[8:], dtype=np.uint8)
labels[np.arange(num_images), indices] += 1
return labels
def read_mnist_images(fn):
with gzip.open(fn, 'rb') as f:
content = f.read()
num_images = int.from_bytes(content[4:8], byteorder='big')
height = int.from_bytes(content[8:12], byteorder='big')
width = int.from_bytes(content[12:16], byteorder='big')
images = np.frombuffer(
content[16:], dtype=np.uint8,
).reshape((num_images, height, width))
images = images.astype(np.float32) / 255.
return images
class ConvNet(ag.layers.Layer):
def __init__(self):
super(ConvNet, self).__init__()
self._conv2d_1_1 = ag.layers.Conv2D(
filters=16,
strides=(1, 1),
kernel_size=(3, 3),
use_bias=False,
padding="SAME",
activation=None,
kernel_initializer="truncated_normal",
)
self._conv2d_1_2 = ag.layers.Conv2D(
filters=32,
strides=(2, 2),
kernel_size=(3, 3),
use_bias=False,
padding="SAME",
activation=None,
kernel_initializer="truncated_normal",
)
self._conv2d_2_1 = ag.layers.Conv2D(
filters=64,
strides=(1, 1),
kernel_size=(3, 3),
use_bias=False,
padding="SAME",
activation=None,
kernel_initializer="truncated_normal",
)
self._conv2d_2_2 = ag.layers.Conv2D(
filters=128,
strides=(2, 2),
kernel_size=(3, 3),
use_bias=False,
padding="SAME",
activation=None,
kernel_initializer="truncated_normal",
)
self._conv2d_3 = ag.layers.Conv2D(
filters=256,
strides=(2, 2),
kernel_size=(3, 3),
use_bias=False,
padding="SAME",
activation=None,
kernel_initializer="truncated_normal",
)
self._bn_1_1 = ag.layers.BatchNormalization(momentum=0.99, epsilon=0.0001)
self._bn_1_2 = ag.layers.BatchNormalization(momentum=0.99, epsilon=0.0001)
self._bn_2_1 = ag.layers.BatchNormalization(momentum=0.99, epsilon=0.0001)
self._bn_2_2 = ag.layers.BatchNormalization(momentum=0.99, epsilon=0.0001)
self._bn_3 = ag.layers.BatchNormalization(momentum=0.99, epsilon=0.0001)
self._dense = ag.layers.Dense(10, use_bias=True)
def __call__(self, inputs, training=False):
outputs_1_1 = ag.relu(
self._bn_1_1(self._conv2d_1_1(inputs), training=training),
)
outputs_1_2 = ag.relu(
self._bn_1_2(self._conv2d_1_2(outputs_1_1), training=training),
)
outputs_2_1 = ag.relu(
self._bn_2_1(self._conv2d_2_1(outputs_1_2), training=training),
)
outputs_2_2 = ag.relu(
self._bn_2_2(self._conv2d_2_2(outputs_2_1), training=training),
)
outputs_3 = ag.relu(
self._bn_3(self._conv2d_3(outputs_2_2), training=training),
)
pool = ag.reduce_mean(outputs_3, [1, 2])
logits = self._dense(pool)
return logits
def minibatch_generator(labels, images, batch_size):
while True:
which = random_state.choice(train_images.shape[0], batch_size, False)
yield labels[which], images[which]
if __name__ == "__main__":
graph = ag.get_default_graph()
batch_size = 50
# build graph
convnet = ConvNet()
inputs = ag.placeholder(shape=(batch_size, 28, 28, 1))
labels = ag.placeholder(shape=(batch_size, 10))
logits = convnet(inputs, True)
preds = convnet(inputs, False)
losses = ag.softmax_cross_entropy_with_logits(labels=labels, logits=logits)
loss = ag.reduce_mean(losses)
variables = convnet.variables
# optimizer
adam = ag.optimizers.AdamOptimizer(
alpha=0.001, beta1=.9, beta2=.999, epsilon=1e-8,
)
grads_and_vars = adam.compute_gradients(loss, variables)
# data
path = "/home/chaoji/data/mnist"
train_images = read_mnist_images(
os.path.join(path, "train-images-idx3-ubyte.gz"),
)
train_labels = read_mnist_labels(
os.path.join(path, "train-labels-idx1-ubyte.gz"),
)
test_images = read_mnist_images(
os.path.join(path, "t10k-images-idx3-ubyte.gz"),
)
test_labels = read_mnist_labels(
os.path.join(path, "t10k-labels-idx1-ubyte.gz"),
)
train_images = train_images[..., np.newaxis]
test_images = test_images[..., np.newaxis]
train_data_generator = minibatch_generator(
train_labels, train_images, batch_size,
)
# training loops
iterations = 1000
for i in np.arange(iterations):
batch_labels, batch_images = next(train_data_generator)
inputs.set_value(batch_images)
labels.set_value(batch_labels)
if i % 100 == 0:
print(f"step: {i}, loss: {loss.eval()}")
adam.apply_gradients(grads_and_vars, reset_runtime=True)
if i % 100 == 0:
assert len(
graph.runtime._values,
) == 0 and len(graph.runtime._placeholder_values) == 0
inputs.set_value(test_images)
print((preds.eval().argmax(axis=1) == test_labels.argmax(axis=1)).mean())
graph.runtime.reset()