forked from xiaohan2012/twitter-sent-dnn
-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_numpy_dcnn.py
147 lines (110 loc) · 3.8 KB
/
test_numpy_dcnn.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
import theano
import numpy as np
from dcnn import DCNN
from logreg import LogisticRegression
from dcnn_train import (WordEmbeddingLayer, ConvFoldingPoolLayer)
#########################
# THEANO PART #
#########################
x_symbol = theano.tensor.imatrix('x') # the word indices matrix
y_symbol = theano.tensor.ivector('y') # the sentiment labels
rng = np.random.RandomState(1234)
vocab_size = 10
embed_dm = 8
embeddings = np.asarray(np.random.rand(vocab_size, embed_dm),
dtype = theano.config.floatX)
layer1 = WordEmbeddingLayer(rng,
input = x_symbol,
vocab_size = vocab_size,
embed_dm = embed_dm,
embeddings = theano.shared(value = embeddings,
name = "embeddings"
)
)
filter_shape = (3, 1, 1, 2)
W = np.asarray(np.random.rand(3, 1, 1, 2),
dtype = theano.config.floatX
)
b = np.asarray(np.random.rand(3),
dtype = theano.config.floatX
)
k = 3
layer2 = ConvFoldingPoolLayer(rng = rng,
input = layer1.output,
filter_shape = filter_shape,
k = k,
fold = 1,
W = theano.shared(value = W, name = "W"),
b = theano.shared(value = b, name = "b")
)
n_in = filter_shape[0] * k * embed_dm / 2
n_out = 5
W_logreg = np.asarray(np.random.rand(n_in, n_out),
dtype = theano.config.floatX)
b_logreg = np.asarray(np.random.rand(n_out),
dtype = theano.config.floatX)
layer3 = LogisticRegression(rng = rng,
input = layer2.output.flatten(2),
n_in = n_in,
n_out = n_out,
W = theano.shared(value = W_logreg, name = "W_logreg"),
b = theano.shared(value = b_logreg, name = "b_logreg")
)
f1 = theano.function(inputs = [x_symbol, y_symbol],
outputs = layer3.nnl(y_symbol)
)
f2 = theano.function(inputs = [x_symbol, y_symbol],
outputs = layer3.errors(y_symbol)
)
f3 = theano.function(inputs = [x_symbol],
outputs = layer3.p_y_given_x
)
f_el = theano.function(inputs = [x_symbol],
outputs = layer1.output
)
f_cl = theano.function(inputs = [x_symbol],
outputs = layer2.output
)
#########################
# NUMPY PART #
#########################
class Params(object):
pass
p = Params()
p.embeddings = embeddings
p.conv_layer_n = 1
p.ks = [3]
p.fold = [1]
p.W = [W]
p.b = [b]
p.W_logreg = W_logreg
p.b_logreg = b_logreg
dcnn = DCNN(p)
##################### Testing ####################
from test_util import (assert_matrix_eq, assert_about_eq)
x = np.asarray(np.random.randint(vocab_size, size = (3, 6)),
dtype=np.int32
)
y = np.asarray(np.random.randint(5, size = 3),
dtype=np.int32
)
########### Embedding layer ##############
actual = f_el(x)
expected = dcnn.e_layer.output(x)
assert_matrix_eq(actual, expected, "Embedding")
########## Conv layer ###################
actual = dcnn._c_layer_output(x)
expected = f_cl(x)
assert_matrix_eq(actual, expected, "Conv")
########## Output layer ###################
actual = dcnn._p_y_given_x(x)
expected = f3(x)
assert_matrix_eq(actual, expected, "p_y_given_x")
########## errors ###########
actual = dcnn._errors(x, y)
expected = f2(x, y)
assert_about_eq(actual, expected, "errors")
########## nnl ###########
actual = dcnn._nnl(x, y)
expected = f1(x, y)
assert_about_eq(actual, expected, "nnl")