/
a08_predict_ensemble.py
232 lines (223 loc) · 16.4 KB
/
a08_predict_ensemble.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
# -*- coding: utf-8 -*-
#prediction using multi-models. take out: create multiple graphs. each graph associate with a session. add logits of models.
#process--->1.load data(X:list of lint,y:int). 2.create session. 3.feed data. 4.predict
import sys
reload(sys)
sys.setdefaultencoding('utf8')
import tensorflow as tf
import numpy as np
import os
from a3_entity_network import EntityNetwork
sys.path.append("..")
from a08_DynamicMemoryNetwork.data_util_zhihu import load_data_predict,load_final_test_data,create_voabulary,create_voabulary_label
from tflearn.data_utils import pad_sequences #to_categorical
import codecs
from a08_DynamicMemoryNetwork.a8_dynamic_memory_network import DynamicMemoryNetwork
from p7_TextCNN_model import TextCNN
from p71_TextRCNN_mode2 import TextRCNN
#configuration
FLAGS=tf.app.flags.FLAGS
tf.app.flags.DEFINE_integer("num_classes",1999,"number of label")
tf.app.flags.DEFINE_float("learning_rate",0.01,"learning rate")
tf.app.flags.DEFINE_integer("batch_size", 80, "Batch size for training/evaluating.") #批处理的大小 32-->128
tf.app.flags.DEFINE_integer("decay_steps", 6000, "how many steps before decay learning rate.") #6000批处理的大小 32-->128
tf.app.flags.DEFINE_float("decay_rate", 1.0, "Rate of decay for learning rate.") #0.65一次衰减多少
tf.app.flags.DEFINE_string("ckpt_dir_dmn","../checkpoint_dynamic_memory_network/","checkpoint location for the model")
tf.app.flags.DEFINE_integer("sequence_length",60,"max sentence length")
tf.app.flags.DEFINE_integer("embed_size",100,"embedding size")
tf.app.flags.DEFINE_boolean("is_training",False,"is traning.true:tranining,false:testing/inference")
tf.app.flags.DEFINE_integer("num_epochs",1,"number of epochs to run.")
tf.app.flags.DEFINE_integer("validate_every", 1, "Validate every validate_every epochs.") #每10轮做一次验证
tf.app.flags.DEFINE_boolean("use_embedding",True,"whether to use embedding or not.")
#tf.app.flags.DEFINE_string("cache_path","text_cnn_checkpoint/data_cache.pik","checkpoint location for the model")
tf.app.flags.DEFINE_string("traning_data_path","../train-zhihu4-only-title-all.txt","path of traning data.") #O.K.train-zhihu4-only-title-all.txt-->training-data/test-zhihu4-only-title.txt--->'training-data/train-zhihu5-only-title-multilabel.txt'
tf.app.flags.DEFINE_string("word2vec_model_path","../zhihu-word2vec-title-desc.bin-100","word2vec's vocabulary and vectors") #zhihu-word2vec.bin-100-->zhihu-word2vec-multilabel-minicount15.bin-100
tf.app.flags.DEFINE_boolean("multi_label_flag",True,"use multi label or single label.")
tf.app.flags.DEFINE_integer("hidden_size",100,"hidden size")
tf.app.flags.DEFINE_string("predict_target_file","zhihu_result_ensemble_2_0814.csv","target file path for final prediction")
tf.app.flags.DEFINE_string("predict_source_file",'../test-zhihu-forpredict-title-desc-v6.txt',"target file path for final prediction") #test-zhihu-forpredict-v4only-title.txt
tf.app.flags.DEFINE_integer("story_length",1,"story length")
tf.app.flags.DEFINE_boolean("use_gated_gru",True,"whether to use gated gru as memory update mechanism. if false,use weighted sum of candidate sentences according to gate")
tf.app.flags.DEFINE_integer("num_pass",3,"number of pass to run") #e.g. num_pass=1,2,3,4.
tf.app.flags.DEFINE_float("l2_lambda", 0.0001, "l2 regularization")
tf.app.flags.DEFINE_boolean("decode_with_sequences",False,"if your task is sequence generating, you need to set this true.default is false, for predict a label")
###################################above from dynamic memory. below from entityNet#######################################################################################
tf.app.flags.DEFINE_string("ckpt_dir_entity","../checkpoint_entity_network5-b40-60-l2B/","checkpoint location for the model")
tf.app.flags.DEFINE_integer("block_size",40,"block size")
tf.app.flags.DEFINE_boolean("use_bi_lstm",True,"whether to use bi-directional lstm for encode of story and query")
###################################above from dynamic memory. below from entityNet#######################################################################################
tf.app.flags.DEFINE_string("ckpt_dir_cnn","../checkpoint_text_cnn/text_cnn_title_desc_checkpoint_exp512/bak_important/","checkpoint location for the model")
tf.app.flags.DEFINE_integer("sentence_len",100,"max sentence length")
tf.app.flags.DEFINE_integer("num_filters", 512, "number of filters") #128
filter_sizes=[3,4,5,7,10,15,20,25]
###################################above is TextRCNN######################################################################################################################
tf.app.flags.DEFINE_string("ckpt_dir_rcnn","../checkpoint_rcnn/text_rcnn_title_desc_checkpoint2/","checkpoint location for the model")
#tf.app.flags.DEFINE_integer("sentence_len",100,"max sentence length")
###################################above is RCNN############################################################################################################################
###################################above is TextCNN_256embedding############################################################################################################
tf.app.flags.DEFINE_string("ckpt_dir_cnn_256_embedding","../checkpoint_text_cnn/text_cnn_title_desc_checkpoint_exp512_0814/","checkpoint location for the model")
filter_sizes_256_embedding=[3,4,5,6,7,8,9,10,15,20,25] #[1,2,3,4,5,6,7,8,9,10]#[1,2,3,4,5,6,7,8,9]#[5,6,7,8,9] #[2,3,5,6,7,8]#[3,4,5,7,10,15,20,25] #[1,2,3,4,5,6,7][3,5,7]#[7,8,9,10,15,20,25] #[3,4,5,7,10,15,20,25]-->[6,7,8,10,15,20,25,30,35]BAD EPOCH2:13.2 #
tf.app.flags.DEFINE_integer("num_filters_256_embedding", 128, "number of filters") #256--->512--->600
tf.app.flags.DEFINE_integer("embed_size_256_embedding", 256, "embedding and hidden size") #256--->512--->600
###################################above is TextCNN_256embedding############################################################################################################
###################################above is HAN############################################################################################################
tf.app.flags.DEFINE_string("ckpt_dir_cnn_256_embedding","../checkpoint_text_cnn/text_cnn_title_desc_checkpoint_exp512_0814/","checkpoint location for the model")
###################################above is THAN############################################################################################################
def main(_):
# 1.load data with vocabulary of words and labels
vocabulary_word2index, vocabulary_index2word = create_voabulary(word2vec_model_path=FLAGS.word2vec_model_path,name_scope="dynamic_memory_network")
vocab_size = len(vocabulary_word2index)
vocabulary_word2index_label, vocabulary_index2word_label = create_voabulary_label(name_scope="dynamic_memory_network")
questionid_question_lists=load_final_test_data(FLAGS.predict_source_file)
test= load_data_predict(vocabulary_word2index,vocabulary_word2index_label,questionid_question_lists)
testX=[]
question_id_list=[]
for tuple in test:
question_id,question_string_list=tuple
question_id_list.append(question_id)
testX.append(question_string_list)
# 2.Data preprocessing: Sequence padding
print("start padding....")
testX2 = pad_sequences(testX, maxlen=FLAGS.sequence_length, value=0.) # padding to max length
testX2_cnn = pad_sequences(testX, maxlen=FLAGS.sentence_len, value=0.) # padding to max length, for CNN
print("end padding...")
# 3.create session.
config=tf.ConfigProto()
config.gpu_options.allow_growth=True
graph1 = tf.Graph().as_default()
graph2 = tf.Graph().as_default()
graph3 = tf.Graph().as_default()
graph4 = tf.Graph().as_default()
graph5 = tf.Graph().as_default()
global sess_dmn
global sess_entity
global sess_cnn
global sess_rcnn
with graph1:#DynamicMemoryNetwork
sess_dmn = tf.Session(config=config)
model_dmn = DynamicMemoryNetwork(FLAGS.num_classes, FLAGS.learning_rate, FLAGS.batch_size, FLAGS.decay_steps, FLAGS.decay_rate, FLAGS.sequence_length,
FLAGS.story_length,vocab_size, FLAGS.embed_size, FLAGS.hidden_size, FLAGS.is_training,num_pass=FLAGS.num_pass,
use_gated_gru=FLAGS.use_gated_gru,decode_with_sequences=FLAGS.decode_with_sequences,multi_label_flag=FLAGS.multi_label_flag,l2_lambda=FLAGS.l2_lambda)
saver_dmn = tf.train.Saver()
if os.path.exists(FLAGS.ckpt_dir_dmn + "checkpoint"):
print("Restoring Variables from Checkpoint of DMN.")
saver_dmn.restore(sess_dmn, tf.train.latest_checkpoint(FLAGS.ckpt_dir_dmn))
else:
print("Can't find the checkpoint.going to stop.DMN")
return
with graph2:#EntityNet
sess_entity = tf.Session(config=config)
model_entity = EntityNetwork(FLAGS.num_classes, FLAGS.learning_rate, FLAGS.batch_size, FLAGS.decay_steps, FLAGS.decay_rate, FLAGS.sequence_length,
FLAGS.story_length,vocab_size, FLAGS.embed_size, FLAGS.hidden_size, FLAGS.is_training,
multi_label_flag=True, block_size=FLAGS.block_size,use_bi_lstm=FLAGS.use_bi_lstm)
saver_entity = tf.train.Saver()
if os.path.exists(FLAGS.ckpt_dir_entity + "checkpoint"):
print("Restoring Variables from Checkpoint of EntityNet.")
saver_entity.restore(sess_entity, tf.train.latest_checkpoint(FLAGS.ckpt_dir_entity))
else:
print("Can't find the checkpoint.going to stop.EntityNet.")
return
with graph3:#TextCNN
sess_cnn=tf.Session(config=config)
model_cnn = TextCNN(filter_sizes, FLAGS.num_filters, FLAGS.num_classes, FLAGS.learning_rate, FLAGS.batch_size,
FLAGS.decay_steps, FLAGS.decay_rate,FLAGS.sentence_len, vocab_size, FLAGS.embed_size, FLAGS.is_training)
saver_cnn = tf.train.Saver()
if os.path.exists(FLAGS.ckpt_dir_cnn + "checkpoint"):
print("Restoring Variables from Checkpoint.TextCNN.")
saver_cnn.restore(sess_cnn, tf.train.latest_checkpoint(FLAGS.ckpt_dir_cnn))
else:
print("Can't find the checkpoint.going to stop.TextCNN.")
return
with graph5: #TextCNN_256embedding
sess_cnn_256_embedding = tf.Session(config=config)
model_cnn_256_embedding = TextCNN(filter_sizes_256_embedding, FLAGS.num_filters_256_embedding, FLAGS.num_classes, FLAGS.learning_rate,
FLAGS.batch_size,FLAGS.decay_steps, FLAGS.decay_rate, FLAGS.sentence_len, vocab_size,
FLAGS.embed_size_256_embedding, FLAGS.is_training)
saver_cnn_256_embedding = tf.train.Saver()
if os.path.exists(FLAGS.ckpt_dir_cnn_256_embedding + "checkpoint"):
print("Restoring Variables from Checkpoint.TextCNN_256_embedding")
saver_cnn_256_embedding.restore(sess_cnn_256_embedding, tf.train.latest_checkpoint(FLAGS.ckpt_dir_cnn_256_embedding))
else:
print("Can't find the checkpoint.going to stop.TextCNN_256_embedding.")
return
#with graph4:#RCNN
# sess_rcnn=tf.Session(config=config)
# model_rcnn=TextRCNN(FLAGS.num_classes, FLAGS.learning_rate, FLAGS.decay_steps, FLAGS.decay_rate,FLAGS.sentence_len,
# vocab_size,FLAGS.embed_size,FLAGS.is_training,FLAGS.batch_size,multi_label_flag=FLAGS.multi_label_flag)
# saver_rcnn = tf.train.Saver()
# if os.path.exists(FLAGS.ckpt_dir_rcnn + "checkpoint"):
# print("Restoring Variables from Checkpoint.TextRCNN.")
# saver_rcnn.restore(sess_rcnn, tf.train.latest_checkpoint(FLAGS.ckpt_dir_rcnn))
# else:
# print("Can't find the checkpoint.going to stop.TextRCNN.")
# return
# 5.feed data, to get logits
number_of_training_data=len(testX2);print("number_of_training_data:",number_of_training_data)
index=0
predict_target_file_f = codecs.open(FLAGS.predict_target_file, 'a', 'utf8')
global sess_dmn
global sess_entity
for start, end in zip(range(0, number_of_training_data, FLAGS.batch_size),range(FLAGS.batch_size, number_of_training_data+1, FLAGS.batch_size)):
#1.DMN
logits_dmn=sess_dmn.run(model_dmn.logits,feed_dict={model_dmn.query:testX2[start:end],model_dmn.story: np.expand_dims(testX2[start:end],axis=1),
model_dmn.dropout_keep_prob:1.0})
#2.EntityNet
logits_entity=sess_entity.run(model_entity.logits,feed_dict={model_entity.query:testX2[start:end],model_entity.story: np.expand_dims(testX2[start:end],axis=1),
model_entity.dropout_keep_prob:1.0})
#3.CNN
logits_cnn = sess_cnn.run(model_cnn.logits,feed_dict={model_cnn.input_x: testX2_cnn[start:end], model_cnn.dropout_keep_prob: 1})
#4.RCNN
#logits_rcnn = sess_rcnn.run(model_rcnn.logits, feed_dict={model_rcnn.input_x: testX2_cnn[start:end],model_rcnn.dropout_keep_prob: 1}) # 'shape of logits:', ( 1, 1999)
#5.CN_256_original_embeddding
logits_cnn_256_embedding =sess_cnn_256_embedding.run(model_cnn_256_embedding.logits,feed_dict={model_cnn_256_embedding.input_x: testX2_cnn[start:end],
model_cnn_256_embedding.dropout_keep_prob: 1})
#how to combine to logits: average
logits=logits_cnn*0.3+logits_cnn_256_embedding*0.3+logits_entity*0.2+logits_dmn*0.2#+logits_rcnn*0.15
question_id_sublist=question_id_list[start:end]
get_label_using_logits_batch(question_id_sublist, logits, vocabulary_index2word_label, predict_target_file_f)
index=index+1
predict_target_file_f.close()
# get label using logits
def get_label_using_logits(logits,vocabulary_index2word_label,top_number=5):
index_list=np.argsort(logits)[-top_number:] #print("sum_p", np.sum(1.0 / (1 + np.exp(-logits))))
index_list=index_list[::-1]
label_list=[]
for index in index_list:
label=vocabulary_index2word_label[index]
label_list.append(label) #('get_label_using_logits.label_list:', [u'-3423450385060590478', u'2838091149470021485', u'-3174907002942471215', u'-1812694399780494968', u'6815248286057533876'])
return label_list
# get label using logits
def get_label_using_logits_with_value(logits,vocabulary_index2word_label,top_number=5):
index_list=np.argsort(logits)[-top_number:] #print("sum_p", np.sum(1.0 / (1 + np.exp(-logits))))
index_list=index_list[::-1]
value_list=[]
label_list=[]
for index in index_list:
label=vocabulary_index2word_label[index]
label_list.append(label) #('get_label_using_logits.label_list:', [u'-3423450385060590478', u'2838091149470021485', u'-3174907002942471215', u'-1812694399780494968', u'6815248286057533876'])
value_list.append(logits[index])
return label_list,value_list
# write question id and labels to file system.
def write_question_id_with_labels(question_id,labels_list,f):
labels_string=",".join(labels_list)
f.write(question_id+","+labels_string+"\n")
# get label using logits
def get_label_using_logits_batch(question_id_sublist,logits_batch,vocabulary_index2word_label,f,top_number=5):
#print("get_label_using_logits.shape:", logits_batch.shape) # (10, 1999))=[batch_size,num_labels]===>需要(10,5)
for i,logits in enumerate(logits_batch):
index_list=np.argsort(logits)[-top_number:] #print("sum_p", np.sum(1.0 / (1 + np.exp(-logits))))
index_list=index_list[::-1]
label_list=[]
for index in index_list:
label=vocabulary_index2word_label[index]
label_list.append(label) #('get_label_using_logits.label_list:', [u'-3423450385060590478', u'2838091149470021485', u'-3174907002942471215', u'-1812694399780494968', u'6815248286057533876'])
#print("get_label_using_logits.label_list",label_list)
write_question_id_with_labels(question_id_sublist[i], label_list, f)
f.flush()
#return label_list
# write question id and labels to file system.
def write_question_id_with_labels(question_id,labels_list,f):
labels_string=",".join(labels_list)
f.write(question_id+","+labels_string+"\n")
if __name__ == "__main__":
tf.app.run()