-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
executable file
·185 lines (175 loc) · 7.34 KB
/
train.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
#!/usr/bin/env python
# -*- coding=utf8 -*-
"""
# Author: zhao zhishan
# Created Time : 2021-01-13 07:05
# File Name: train.py
# Description:
"""
import os
import sys
import shutil
import logging
import argparse
from importlib import import_module
import tensorflow as tf
from tensorflow.python.client import timeline
from sklearn.metrics import roc_auc_score, log_loss
from data_loader import build_dataset, create_dataset_iterator, get_vocab_size
def evaluate(data_iter, model, sess):
data_iter_handle = sess.run(data_iter.string_handle())
sess.run(data_iter.initializer)
scores = []
labels = []
while True:
try:
batch_labels, batch_scores = sess.run(
[model.y_true, model.scores],
feed_dict={model.handle:data_iter_handle,
model.is_train: False})
scores.extend(batch_scores)
labels.extend(batch_labels)
except tf.errors.OutOfRangeError:
break
auc = roc_auc_score(labels, scores)
logloss = log_loss(labels, scores, eps=1e-7)
return auc, logloss
def train_and_eval(train_data_iter, val_data_iter, eval_data_iter,
model, sess, eval_step, model_dir, print_each=1000):
step = 0
best_auc = 0
last_logloss = 0
hold_step = 0
best_logloss = sys.maxsize
train_handle = sess.run(train_data_iter.string_handle())
if os.path.exists(model_dir):
shutil.rmtree(model_dir)
os.makedirs(model_dir)
while True:
if step > 0 and step % eval_step == 0:
logging.info('=============start evaluate================')
val_auc, val_logloss = evaluate(val_data_iter, model, sess)
eval_auc, eval_logloss = evaluate(eval_data_iter, model, sess)
if eval_logloss < best_logloss:
#if eval_auc > best_auc:
best_logloss = eval_logloss
best_auc = eval_auc
hold_step = 0
model.saver.save(sess, '{}/model.ckpt'.format(model_dir), global_step=step)
logging.info('step: {}, train auc: {:.4}, train logloss: {:.4}, eval auc: {:.4}, eval logloss: {:.4}, best_auc: {:.4}, best_logloss: {:.4}'\
.format(step, val_auc, val_logloss,
eval_auc, eval_logloss, best_auc, best_logloss))
logging.info('=============end evaluate================')
if hold_step > 6:
logging.info('eval logloss not decrease, early stopping')
break
hold_step += 1
try:
loss, _, lr = sess.run(
[model.loss, model.train_op, model.lr],
feed_dict={model.handle:train_handle,
model.is_train: True})
if step > 0 and step % print_each == 0:
logging.info('step: {}, train logloss: {:.4}, lr: {:.4}'.format(step, loss, lr))
except tf.errors.OutOfRangeError:
break
step += 1
def run(args):
vocab_size = get_vocab_size(args.feature_size_file)
train_dataset = build_dataset(args.train_data, args.epoch, args.batch_size)
val_dataset = build_dataset(args.val_data, 1, args.batch_size*5)
eval_dataset = build_dataset(args.eval_data, 1, args.batch_size*5)
test_dataset = build_dataset(args.test_data, 1, args.batch_size*5)
handle, batch_data, train_data_iter, val_data_iter, eval_data_iter, test_data_iter =\
create_dataset_iterator(train_dataset, val_dataset, eval_dataset, test_dataset)
y_true, feat_idx, feat_val = batch_data
params = {'lr': args.learning_rate,
'l2_reg': args.l2_reg,
'dropout_rate': args.dropout_rate,
'batch_size': args.batch_size,
'emb_size': args.emb_size}
logging.info('params: {}'.format(params))
model_module = import_module('models.{}'.format(args.model_name))
model = model_module.Model(vocab_size, args.field_num, params)
model.init_graph(y_true, feat_idx, feat_val, handle)
session_conf = tf.ConfigProto()
session_conf.gpu_options.allow_growth = True
with tf.Session(config=session_conf) as sess:
sess.run(tf.global_variables_initializer())
logging.info('----model running----')
train_and_eval(
train_data_iter, val_data_iter, eval_data_iter, model, sess, args.eval_step, args.model_dir)
ckpt = tf.train.latest_checkpoint(args.model_dir)
model.saver.restore(sess, ckpt)
test_auc, test_logloss = evaluate(test_data_iter, model, sess)
logging.info('test auc: {:.4}, test logloss: {:.4}'.format(test_auc, test_logloss))
logging.info('----done----')
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--feature_size_file',
type=str,
required=True,
help='feature size info file')
parser.add_argument('--train_data',
type=str,
required=True,
help='train data path')
parser.add_argument('--val_data',
type=str,
required=True,
help='val data path(select from train data)')
parser.add_argument('--eval_data',
type=str,
required=True,
help='eval data path')
parser.add_argument('--test_data',
type=str,
required=True,
help='test data path')
parser.add_argument('--model_name',
type=str,
required=True,
help='model name')
parser.add_argument('--model_dir',
type=str,
required=True,
help='save model checkpoint directory')
parser.add_argument('--learning_rate', '-lr',
type=float,
required=True)
parser.add_argument('--field_num',
type=int,
required=True)
parser.add_argument('--epoch',
type=int,
required=True)
parser.add_argument('--l2_reg',
type=float,
required=True)
parser.add_argument('--dropout_rate',
type=float,
required=True)
parser.add_argument('--batch_size',
type=int,
required=True)
parser.add_argument('--emb_size',
type=int,
required=True)
parser.add_argument('--eval_step',
type=int,
required=True)
parser.add_argument('--version',
type=str,
required=True)
parser.add_argument('--gpu',
type=str,
required=True)
args = parser.parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu
tf.logging.set_verbosity(tf.logging.INFO)
logging.basicConfig(level=logging.INFO,
filename='./log_{}.txt'.format(args.version),
filemode='w',
format='%(asctime)s - %(filename)s[line:%(lineno)d] - %(levelname)s: %(message)s')
with tf.Graph().as_default():
run(args)