-
Notifications
You must be signed in to change notification settings - Fork 3
/
eval.py
170 lines (139 loc) · 7 KB
/
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
import itertools
import os
import sys
import tensorflow as tf
from tensorflow.keras.utils import Progbar
import numpy as np
from models import mlp
from utils import ModelConfig
from misc import evaluate, evaluate_bmsd, evaluate_lfirst
from logger import QuickLogger
flags = tf.app.flags
FLAGS = flags.FLAGS
# common options
flags.DEFINE_string('root_dir', '/data', 'root directory')
flags.DEFINE_string('output_dir', None, 'output directory')
flags.DEFINE_string('exp_dir', None, 'experiment directory')
flags.DEFINE_boolean('is_quant', False, 'quantization or not')
flags.DEFINE_integer('weight_bits', 8, 'the number of bits of weight')
flags.DEFINE_integer('activation_bits', 8, 'the number of bits of activation')
flags.DEFINE_integer('seed', 2019, 'seed')
flags.DEFINE_integer('val_epoch', None, 'validation epoch')
# options for dataset
flags.DEFINE_enum('tokenizer', 'okt', ['okt', 'mecab', 'whitespace', 'okt_sub'],
'tokenizer used to generate .tfrecord')
flags.DEFINE_integer('vocab_size', 150000, 'embedding size')
flags.DEFINE_integer('max_len', 20, 'maximum word count per product')
flags.DEFINE_string('postfix', '', 'postfix appended to .tfrecord')
flags.DEFINE_integer('total_chunk', 20, 'the number of tfrecords for validation')
flags.DEFINE_integer('val_chunk', 1, 'the number of tfrecords for validation')
flags.DEFINE_integer('batch_size', 1024, 'batch size')
flags.DEFINE_boolean('zero_based', False, 'label is zero-based')
# options for network
flags.DEFINE_enum('model', 'mlp', ['mlp', 'lstm'], 'type of classifier')
flags.DEFINE_boolean('bmsd', False, 'whether bmsd is used or not')
flags.DEFINE_boolean('use_b', False, 'whether b is used or not')
flags.DEFINE_boolean('use_m', False, 'whether m is used or not')
flags.DEFINE_integer('embd_size', 256, 'embedding size')
flags.DEFINE_enum('embd_avg_type', 'divide_by_valid', ['divide_by_valid', 'simple_mean', 'simple_sum'], 'type of averaged embedding')
flags.DEFINE_enum('embd_init_type', 'none', ['none', 'uniform'], 'type of initializer for embedding')
flags.DEFINE_enum('noise_type', 'adv', ['none', 'adv'], 'type of classifier')
flags.DEFINE_float('unigram_adv_noise', 5e-1, 'adv noise strength for unigram')
flags.DEFINE_float('img_adv_noise', 5e0, 'adv noise strength for img feature')
flags.DEFINE_float('embd_dropout_rate', 0.5, 'dropout rate for embedding')
flags.DEFINE_float('fc_dropout_rate', 0.5, 'dropout rate for fc')
flags.DEFINE_integer('fc_img_feat', 0, 'the size of fc layer for image feat')
flags.DEFINE_multi_integer('fc_layers', [4096], 'the size of fc layer for classifiers')
logger = QuickLogger(log_dir=FLAGS.exp_dir, log_name='val_logs.txt').get_logger()
def val(sess, val_vars):
sess.run([val_vars.it.initializer, tf.local_variables_initializer()])
val_progbar = Progbar(None, stateful_metrics=['loss', 'acc', 'num_word',])
preds = []
preds_b = []
preds_m = []
preds_s = []
preds_d = []
gts = []
indices = tf.argmax(val_vars.logits, 1)
feed_dict = [val_vars.loss_metric, val_vars.acc_metric, val_vars.num_word_metric, indices, val_vars.labels]
if FLAGS.bmsd:
unshift = 1 if FLAGS.zero_based else 0
indices_b = tf.argmax(val_vars.logits_b, 1)+unshift
indices_m = tf.argmax(val_vars.logits_m, 1)+unshift
indices_s = tf.argmax(val_vars.logits_s, 1)+unshift
indices_d = tf.argmax(val_vars.logits_d, 1)+unshift
feed_dict.append(indices_b)
feed_dict.append(indices_m)
feed_dict.append(indices_s)
feed_dict.append(indices_d)
try:
step = 0
while True:
if FLAGS.bmsd:
(loss, _), (acc, _), (num_word, _), pred, gt, pred_b, pred_m, pred_s, pred_d = sess.run(feed_dict)
preds_b.append(pred_b)
preds_m.append(pred_m)
preds_s.append(pred_s)
preds_d.append(pred_d)
else:
(loss, _), (acc, _), (num_word, _), pred, gt = sess.run(feed_dict)
val_progbar.update(step, [('loss', loss), ('acc', acc), ('num_word', num_word), ])
preds.append(pred)
gts.append(gt)
except tf.errors.OutOfRangeError:
logger.info('[*] Validation loss: %.4f, acc: %.4f, num_word: %.2f' % (loss, acc, num_word))
if FLAGS.bmsd:
flat_preds = list(itertools.chain(*preds))
flat_gts = list(itertools.chain(*gts))
flat_preds_b = list(itertools.chain(*preds_b))
flat_preds_m = list(itertools.chain(*preds_m))
flat_preds_s = list(itertools.chain(*preds_s))
flat_preds_d = list(itertools.chain(*preds_d))
score, base_scores = evaluate(flat_gts, flat_preds)
print('\n[!] Label Score: %.4f' % (score))
bmsd_score, bmsd_scores = evaluate_bmsd(flat_gts, flat_preds_b, flat_preds_m, flat_preds_s, flat_preds_d)
print('BMSD Score: %.4f' % (bmsd_score))
final_score = 0.0
for r, s1, s2 in zip([1.0,1.2,1.3,1.4], base_scores, bmsd_scores):
final_score += r * max(s1, s2)
final_score /= 4.0
print('[!] Final Score: %.4f' % (final_score))
lfirst_score, lfirst_scores = evaluate_lfirst(flat_gts, flat_preds, flat_preds_b, flat_preds_m, flat_preds_s, flat_preds_d)
print('[!] Label First Score: %.4f' % (lfirst_score))
else:
flat_preds = list(itertools.chain(*preds))
flat_gts = list(itertools.chain(*gts))
score, base_scores = evaluate(flat_gts, flat_preds)
print('\n[!] Label Score: %.4f' % (score))
return score
def main(unused_argv):
tf.logging.set_verbosity(tf.logging.INFO)
np.random.seed(FLAGS.seed)
tf.gfile.MakeDirs(FLAGS.exp_dir)
DATA_DIR = os.path.join(FLAGS.root_dir, "output_%s" % FLAGS.tokenizer)
TFRECORD_FORMAT = DATA_DIR + ("/train%s" % (FLAGS.postfix)) + ".%02d.tfrecord"
val_files = [TFRECORD_FORMAT % i for i in range(FLAGS.total_chunk-FLAGS.val_chunk+1, FLAGS.total_chunk+1)] # 20
model_config = ModelConfig(FLAGS, train=False)
logger.info('[!] FLAGS.*')
logger.info(FLAGS.flag_values_dict())
logger.info('[!] model_config.*')
logger.info(model_config)
print(model_config)
with tf.Graph().as_default() as graph:
tf.set_random_seed(FLAGS.seed)
val_vars = mlp.Classifier(val_files, train=False, test=True, config=model_config)
saver = tf.train.Saver(max_to_keep=30)
ckpt_path = os.path.join(FLAGS.exp_dir, "model.ckpt-%d" % (FLAGS.val_epoch))
config=tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
saver.restore(sess, ckpt_path)
logger.info('[!] Validation...')
val(sess, val_vars)
if __name__ == '__main__':
flags.mark_flag_as_required('output_dir')
flags.mark_flag_as_required('exp_dir')
flags.mark_flag_as_required('model')
flags.mark_flag_as_required('val_epoch')
#tf.app.run()
main(sys.argv[1])