-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathresnet_cifar_eval.py
176 lines (151 loc) · 6.34 KB
/
resnet_cifar_eval.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
# Copyright 2016 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""ResNet Train/Eval module.
"""
import time
import six
import sys
import cifar_input
import numpy as np
import resnet_model
import tensorflow as tf
FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string('dataset', 'cifar10', 'cifar10 or cifar100.')
tf.app.flags.DEFINE_string('mode', 'train', 'train or eval.')
tf.app.flags.DEFINE_string('train_data_path', '',
'Filepattern for training data.')
tf.app.flags.DEFINE_string('eval_data_path', '',
'Filepattern for eval data')
tf.app.flags.DEFINE_integer('image_size', 32, 'Image side length.')
tf.app.flags.DEFINE_string('train_dir', '',
'Directory to keep training outputs.')
tf.app.flags.DEFINE_string('eval_dir', '',
'Directory to keep eval outputs.')
tf.app.flags.DEFINE_integer('eval_batch_count', 50,
'Number of batches to eval.')
tf.app.flags.DEFINE_bool('eval_once', False,
'Whether evaluate the model only once.')
tf.app.flags.DEFINE_string('log_root', '',
'Directory to keep the checkpoints. Should be a '
'parent directory of FLAGS.train_dir/eval_dir.')
tf.app.flags.DEFINE_integer('num_gpus', 0,
'Number of gpus used for training. (0 or 1)')
tf.flags.DEFINE_string('data_format', 'channels_first',
'channels_first for cuDNN, channels_last for MKL')
tf.flags.DEFINE_integer("num_intra_threads", 0,
"Number of threads to use for intra-op parallelism. If set"
"to 0, the system will pick an appropriate number.")
tf.flags.DEFINE_integer("num_inter_threads", 0,
"Number of threads to use for inter-op parallelism. If set"
"to 0, the system will pick an appropriate number.")
def create_config_proto():
"""Returns session config proto.
Args:
params: Params tuple, typically created by make_params or
make_params_from_flags.
"""
config = tf.ConfigProto()
config.allow_soft_placement = True
if(FLAGS.num_intra_threads != 0):
config.intra_op_parallelism_threads = FLAGS.num_intra_threads
if(FLAGS.num_inter_threads != 0):
config.inter_op_parallelism_threads = FLAGS.num_inter_threads
# config.gpu_options.force_gpu_compatible = params.force_gpu_compatible
# if params.gpu_memory_frac_for_testing > 0:
# config.gpu_options.per_process_gpu_memory_fraction = (
# params.gpu_memory_frac_for_testing)
# if params.xla:
# config.graph_options.optimizer_options.global_jit_level = (
# tf.OptimizerOptions.ON_1)
# if params.enable_layout_optimizer:
# config.graph_options.rewrite_options.layout_optimizer = (
# rewriter_config_pb2.RewriterConfig.ON)
return config
def evaluate(hps):
"""Eval loop."""
batch_size = 100
images, labels = cifar_input.build_input(
FLAGS.dataset, FLAGS.eval_data_path, batch_size, FLAGS.mode)
model = resnet_model.ResNet(hps, images, labels, FLAGS.mode)
model.build_graph(False)
saver = tf.train.Saver()
summary_writer = tf.summary.FileWriter(FLAGS.eval_dir)
sess = tf.Session(config=create_config_proto())
tf.train.start_queue_runners(sess)
best_precision = 0.0
while True:
try:
ckpt_state = tf.train.get_checkpoint_state(FLAGS.log_root)
except tf.errors.OutOfRangeError as e:
tf.logging.error('Cannot restore checkpoint: %s', e)
continue
if not (ckpt_state and ckpt_state.model_checkpoint_path):
tf.logging.info('No model to eval yet at %s', FLAGS.log_root)
continue
tf.logging.info('Loading checkpoint %s', ckpt_state.model_checkpoint_path)
saver.restore(sess, ckpt_state.model_checkpoint_path)
total_prediction, correct_prediction = 0, 0
for _ in six.moves.range(FLAGS.eval_batch_count):
(summaries, loss, predictions, truth, train_step) = sess.run(
[model.summaries, model.cost, model.predictions,
model.labels, model.global_step])
truth = np.argmax(truth, axis=1)
predictions = np.argmax(predictions, axis=1)
correct_prediction += np.sum(truth == predictions)
total_prediction += predictions.shape[0]
precision = 1.0 * correct_prediction / total_prediction
best_precision = max(precision, best_precision)
precision_summ = tf.Summary()
precision_summ.value.add(
tag='Precision', simple_value=precision)
summary_writer.add_summary(precision_summ, train_step)
best_precision_summ = tf.Summary()
best_precision_summ.value.add(
tag='Best Precision', simple_value=best_precision)
summary_writer.add_summary(best_precision_summ, train_step)
summary_writer.add_summary(summaries, train_step)
tf.logging.info('loss: %.3f, precision: %.3f, best precision: %.3f' %
(loss, precision, best_precision))
summary_writer.flush()
if FLAGS.eval_once:
break
time.sleep(60)
def main(_):
if FLAGS.num_gpus == 0:
dev = '/cpu:0'
elif FLAGS.num_gpus == 1:
dev = '/gpu:0'
else:
raise ValueError('Only support 0 or 1 gpu.')
if FLAGS.mode == 'train':
batch_size = 128
elif FLAGS.mode == 'eval':
batch_size = 100
if FLAGS.dataset == 'cifar10':
num_classes = 10
elif FLAGS.dataset == 'cifar100':
num_classes = 100
hps = resnet_model.HParams(num_classes=num_classes,
lrn_rate=0.1,
weight_decay_rate=0.0002,
optimizer='mom')
with tf.device(dev):
if FLAGS.mode == 'train':
train(hps)
elif FLAGS.mode == 'eval':
evaluate(hps)
if __name__ == '__main__':
tf.logging.set_verbosity(tf.logging.INFO)
tf.app.run()