forked from Lapis-Hong/wide_deep
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
253 lines (231 loc) · 11.2 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
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
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Author: lapis-hong
# @Date : 2018/1/15
"""Training Wide and Deep Model."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from __future__ import unicode_literals
import argparse
import os
import shutil
import sys
import time
import tensorflow as tf
from lib.read_conf import Config
from lib.dataset import input_fn
from lib.build_estimator import build_estimator, build_custom_estimator
from lib.utils.util import elapse_time, list_files
CONFIG = Config().train
parser = argparse.ArgumentParser(description='Train Wide and Deep Model.')
parser.add_argument(
'--model_dir', type=str, default=CONFIG["model_dir"],
help='Base directory for the model.')
parser.add_argument(
'--model_type', type=str, default=CONFIG["model_type"],
help="Valid model types: {'wide', 'deep', 'wide_deep'}.")
parser.add_argument(
'--train_epochs', type=int, default=CONFIG["train_epochs"],
help='Number of training epochs.')
parser.add_argument(
'--epochs_per_eval', type=int, default=CONFIG["epochs_per_eval"],
help='The number of training epochs to run between evaluations.')
parser.add_argument(
'--batch_size', type=int, default=CONFIG["batch_size"],
help='Number of examples per batch.')
parser.add_argument(
'--train_data', type=str, default=CONFIG["train_data"],
help='Path to the train data.')
parser.add_argument(
'--eval_data', type=str, default=CONFIG["eval_data"],
help='Path to the validation data.')
parser.add_argument(
'--test_data', type=str, default=CONFIG["test_data"],
help='Path to the test data.')
parser.add_argument(
'--image_train_data', type=str, default=CONFIG["image_train_data"],
help='Path to the train data.')
parser.add_argument(
'--image_eval_data', type=str, default=CONFIG["image_eval_data"],
help='Path to the train data.')
parser.add_argument(
'--image_test_data', type=str, default=CONFIG["image_test_data"],
help='Path to the train data.')
parser.add_argument(
'--keep_train', type=int, default=CONFIG["keep_train"],
help='Whether to keep training on previous trained model.')
def train_and_eval(model):
for n in range(FLAGS.train_epochs):
tf.logging.info('=' * 30 + ' START EPOCH {} '.format(n + 1) + '=' * 30 + '\n')
train_data_list = list_files(FLAGS.train_data) # dir to file list
for f in train_data_list:
t0 = time.time()
tf.logging.info('<EPOCH {}>: Start training {}'.format(n + 1, f))
model.train(
input_fn=lambda: input_fn(f, FLAGS.image_train_data, 'train', FLAGS.batch_size),
hooks=None,
steps=None,
max_steps=None,
saving_listeners=None)
tf.logging.info('<EPOCH {}>: Finish training {}, take {} mins'.format(n + 1, f, elapse_time(t0)))
print('-' * 80)
tf.logging.info('<EPOCH {}>: Start evaluating {}'.format(n + 1, FLAGS.eval_data))
t0 = time.time()
results = model.evaluate(
input_fn=lambda: input_fn(FLAGS.eval_data, FLAGS.image_eval_data, 'eval', FLAGS.batch_size),
steps=None, # Number of steps for which to evaluate model.
hooks=None,
checkpoint_path=None, # latest checkpoint in model_dir is used.
name=None)
tf.logging.info('<EPOCH {}>: Finish evaluation {}, take {} mins'.format(n + 1, FLAGS.eval_data, elapse_time(t0)))
print('-' * 80)
# Display evaluation metrics
for key in sorted(results):
print('{}: {}'.format(key, results[key]))
# every epochs_per_eval test the model (use larger test dataset)
if (n+1) % FLAGS.epochs_per_eval == 0:
tf.logging.info('<EPOCH {}>: Start testing {}'.format(n + 1, FLAGS.test_data))
results = model.evaluate(
input_fn=lambda: input_fn(FLAGS.test_data, FLAGS.image_test_data, 'pred', FLAGS.batch_size),
steps=None, # Number of steps for which to evaluate model.
hooks=None,
checkpoint_path=None, # If None, the latest checkpoint in model_dir is used.
name=None)
tf.logging.info('<EPOCH {}>: Finish testing {}, take {} mins'.format(n + 1, FLAGS.test_data, elapse_time(t0)))
print('-' * 80)
# Display evaluation metrics
for key in sorted(results):
print('{}: {}'.format(key, results[key]))
def dynamic_train(model):
"""Dynamic train mode.
For example:
train_data_files: [0301, 0302, 0303, ...]
train mode:
first take 0301 as train data, 0302 as test data;
then keep training take 0302 as train data, 0303 as test data ...
"""
data_files = list_files(FLAGS.train_data)
data_files.sort()
assert len(data_files) > 1, 'Dynamic train mode need more than 1 data file'
for i in range(len(data_files)-1):
train_data = data_files[i]
test_data = data_files[i+1]
tf.logging.info('=' * 30 + ' START TRAINING DATA: {} '.format(train_data) + '=' * 30 + '\n')
for n in range(FLAGS.train_epochs):
t0 = time.time()
tf.logging.info('START TRAIN DATA <{}> <EPOCH {}>'.format(train_data, n + 1))
model.train(
input_fn=lambda: input_fn(train_data, FLAGS.image_train_data, 'train', FLAGS.batch_size),
hooks=None,
steps=None,
max_steps=None,
saving_listeners=None)
tf.logging.info('FINISH TRAIN DATA <{}> <EPOCH {}> take {} mins'.format(train_data, n + 1, elapse_time(t0)))
print('-' * 80)
tf.logging.info('START EVALUATE TEST DATA <{}> <EPOCH {}>'.format(test_data, n + 1))
t0 = time.time()
results = model.evaluate(
input_fn=lambda: input_fn(test_data, FLAGS.image_eval_data, 'eval', FLAGS.batch_size),
steps=None, # Number of steps for which to evaluate model.
hooks=None,
checkpoint_path=None, # latest checkpoint in model_dir is used.
name=None)
tf.logging.info('FINISH EVALUATE TEST DATA <{}> <EPOCH {}>: take {} mins'.format(test_data, n + 1, elapse_time(t0)))
print('-' * 80)
# Display evaluation metrics
for key in sorted(results):
print('{}: {}'.format(key, results[key]))
def train(model):
for n in range(FLAGS.train_epochs):
tf.logging.info('=' * 30 + ' START EPOCH {} '.format(n + 1) + '=' * 30 + '\n')
train_data_list = list_files(FLAGS.train_data) # dir to file list
for f in train_data_list:
t0 = time.time()
tf.logging.info('<EPOCH {}>: Start training {}'.format(n + 1, f))
model.train(
input_fn=lambda: input_fn(f, FLAGS.image_train_data, 'train', FLAGS.batch_size),
hooks=None,
steps=None,
max_steps=None,
saving_listeners=None)
tf.logging.info('<EPOCH {}>: Finish training {}, take {} mins'.format(n + 1, f, elapse_time(t0)))
def train_and_eval_api(model):
train_spec = tf.estimator.TrainSpec(input_fn=lambda: input_fn(FLAGS.train_data, FLAGS.image_train_data, FLAGS.batch_size), max_steps=10000)
eval_spec = tf.estimator.EvalSpec(input_fn=lambda: input_fn(FLAGS.eval_data, FLAGS.image_eval_data, FLAGS.batch_size))
tf.estimator.train_and_evaluate(model, train_spec, eval_spec)
def main(unused_argv):
CONFIG = Config()
print("Using TensorFlow Version %s" % tf.__version__)
assert "1.4" <= tf.__version__, "Need TensorFlow r1.4 or Later."
print('\nModel Type: {}'.format(FLAGS.model_type))
model_dir = os.path.join(FLAGS.model_dir, FLAGS.model_type)
print('\nModel Directory: {}'.format(model_dir))
print("\nUsing Train Config:")
for k, v in CONFIG.train.items():
print('{}: {}'.format(k, v))
print("\nUsing Model Config:")
for k, v in CONFIG.model.items():
print('{}: {}'.format(k, v))
if not FLAGS.keep_train:
# Clean up the model directory if not keep training
shutil.rmtree(model_dir, ignore_errors=True)
print('Remove model directory: {}'.format(model_dir))
# model = build_estimator(model_dir, FLAGS.model_type)
model = build_custom_estimator(model_dir, FLAGS.model_type)
tf.logging.info('Build estimator: {}'.format(model))
if CONFIG.train['dynamic_train']:
train_fn = dynamic_train
print("Using dynamic train mode.")
else:
train_fn = train_and_eval
if CONFIG.distribution["is_distribution"]:
print("Using PID: {}".format(os.getpid()))
cluster = CONFIG.distribution["cluster"]
job_name = CONFIG.distribution["job_name"]
task_index = CONFIG.distribution["task_index"]
print("Using Distributed TensorFlow. Local host: {} Job_name: {} Task_index: {}"
.format(cluster[job_name][task_index], job_name, task_index))
cluster = tf.train.ClusterSpec(CONFIG.distribution["cluster"])
server = tf.train.Server(cluster,
job_name=job_name,
task_index=task_index)
# distributed can not including eval.
train_fn = train
if job_name == 'ps':
# wait for incoming connection forever
server.join()
# sess = tf.Session(server.target)
# queue = create_done_queue(task_index, num_workers)
# for i in range(num_workers):
# sess.run(queue.dequeue())
# print("ps {} received worker {} done".format(task_index, i)
# print("ps {} quitting".format(task_index))
else: # TODO：supervisor & MonotoredTrainingSession & experiment (deprecated)
train_fn(model)
# train_and_eval(model)
# Each worker only needs to contact the PS task(s) and the local worker task.
# config = tf.ConfigProto(device_filters=[
# '/job:ps', '/job:worker/task:%d' % arguments.task_index])
# with tf.device(tf.train.replica_device_setter(
# worker_device="/job:worker/task:%d" % task_index,
# cluster=cluster)):
# e = _create_experiment_fn()
# e.train_and_evaluate() # call estimator's train() and evaluate() method
# hooks = [tf.train.StopAtStepHook(last_step=10000)]
# with tf.train.MonitoredTrainingSession(
# master=server.target,
# is_chief=(task_index == 0),
# checkpoint_dir=args.model_dir,
# hooks=hooks) as mon_sess:
# while not mon_sess.should_stop():
# # mon_sess.run()
# classifier.fit(input_fn=train_input_fn, steps=1)
else:
# local run
train_fn(model)
if __name__ == '__main__':
# Set to INFO for tracking training, default is WARN. ERROR for least messages
tf.logging.set_verbosity(tf.logging.INFO)
FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)