-
Notifications
You must be signed in to change notification settings - Fork 37
/
Product2VecSkipGram.py
168 lines (136 loc) · 6.77 KB
/
Product2VecSkipGram.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
import warnings
from concurrent.futures import ThreadPoolExecutor
from functools import reduce
import tensorflow as tf
import numpy as np
import math
from sklearn.utils import shuffle
class Product2VecSkipGram:
def __init__(self, data, cv_data, batch_size, num_skips, skip_window, vocabulary_size, embedding_size=32,
num_negative_sampled=64, len_ratio = 0.5):
self.data = data
self.cv_data = cv_data
self.data_index = 0
self.batch_size = batch_size
self.num_skips = num_skips
self.skip_window = skip_window
self.embedding_size = embedding_size
self.num_negative_sampled = num_negative_sampled
self.vocabulary_size = vocabulary_size
self.len_ratio = len_ratio
assert batch_size % num_skips == 0
assert num_skips <= 2 * skip_window
self.build_graph()
def predict(self, products):
result = []
for i in range(0, len(products), self.batch_size):
batch = products[i:i+self.batch_size]
batch = self.sess.run(self.gathered, feed_dict={self.train_inputs: batch})
result.append(batch)
return np.concatenate(result, axis=0)
def train(self, num_steps, cv_every_n_steps, cv_steps, lrs):
with ThreadPoolExecutor(max_workers=2) as executor:
average_loss = 0
learning_rate = 1.0
current = executor.submit(self.generate_batch)
for step in range(num_steps):
if step in lrs:
learning_rate = lrs[step]
batch_inputs, batch_labels = current.result()
current = executor.submit(self.generate_batch)
feed_dict = {self.train_inputs: batch_inputs,
self.train_labels: batch_labels,
self.learning_rate: learning_rate}
_, loss_val = self.sess.run([self.optimizer, self.loss], feed_dict=feed_dict)
average_loss += loss_val
if step % 2000 == 0:
if step > 0:
average_loss /= 2000
print('Average loss at step ', step, ': ', average_loss)
average_loss = 0
if step % cv_every_n_steps == 0:
self.data = shuffle(self.data, random_state=0)
self.save_model(step)
cv_loss = 0
for batch_inputs, batch_labels in self.generate_test(cv_steps):
feed_dict = {self.train_inputs: batch_inputs,
self.train_labels: batch_labels,
self.learning_rate: learning_rate}
loss_val = self.sess.run(self.loss, feed_dict=feed_dict)
cv_loss += loss_val
print('CV',cv_loss / cv_steps)
def save_model(self, step):
self.saver.save(self.sess, 'models/prod2vec_skip_gram', global_step=step)
def load_model(self, path):
self.saver.restore(self.sess, path)
def build_graph(self):
self.train_inputs = tf.placeholder(tf.int32, shape=[self.batch_size])
self.train_labels = tf.placeholder(tf.int32, shape=[self.batch_size])
self.learning_rate = tf.placeholder(tf.float32)
# variables
embeddings = tf.Variable(tf.random_uniform([self.vocabulary_size, self.embedding_size], -1.0, 1.0))
softmax_weights = tf.Variable(tf.truncated_normal([self.embedding_size, self.vocabulary_size],
stddev=1.0 / math.sqrt(self.embedding_size)))
softmax_biases = tf.Variable(tf.zeros([self.vocabulary_size]))
self.gathered = tf.gather(embeddings, self.train_inputs)
prediction = tf.matmul(self.gathered, softmax_weights) + softmax_biases
self.loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(labels=self.train_labels, logits=prediction))
self.optimizer = tf.train.GradientDescentOptimizer(self.learning_rate).minimize(self.loss)
self.sess = tf.Session()
self.sess.run(tf.global_variables_initializer())
self.saver = tf.train.Saver()
def inc(self):
self.data_index = (self.data_index + 1) % len(self.data)
def inc_cv(self, data_index):
return (data_index + 1) % len(self.cv_data)
def generate_batch(self):
batch = np.ndarray(shape=(self.batch_size), dtype=np.int32)
labels = np.ndarray(shape=(self.batch_size), dtype=np.int32)
counter = 0
while counter < self.batch_size:
current = self.data.iloc[self.data_index]
if len(current) == 1:
warnings.warn("lenght is one", RuntimeWarning)
self.inc()
continue
span = min(2 * self.skip_window + 1, len(current))
x = target = np.random.randint(0, len(current))
targets_to_avoid = [x]
for j in range(self.num_skips): # target varies!!! X constant!
while target in targets_to_avoid and len(targets_to_avoid) != span:
target = np.random.randint(0, span)
if len(targets_to_avoid) == span or counter == self.batch_size:
break
targets_to_avoid.append(target)
batch[counter] = current[x]
labels[counter] = current[target]
counter += 1
self.inc()
return batch, labels
def generate_test(self, num_steps):
data_index = 0
for _ in range(num_steps):
batch = np.ndarray(shape=(self.batch_size), dtype=np.int32)
labels = np.ndarray(shape=(self.batch_size), dtype=np.int32)
counter = 0
while counter < self.batch_size:
current = self.cv_data.iloc[data_index]
if len(current) == 1:
warnings.warn("lenght is one", RuntimeWarning)
data_index = self.inc_cv(data_index)
continue
span = min(2 * self.skip_window + 1, len(current))
x = target = np.random.randint(0, len(current))
targets_to_avoid = [x]
for j in range(self.num_skips): # target varies!!! X constant!
while target in targets_to_avoid and len(targets_to_avoid) != span:
target = np.random.randint(0, span)
if len(targets_to_avoid) == span or counter == self.batch_size:
break
targets_to_avoid.append(target)
batch[counter] = current[x]
labels[counter] = current[target]
counter += 1
data_index = self.inc_cv(data_index)
yield batch, labels