-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluate.py
197 lines (174 loc) · 6.67 KB
/
evaluate.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
from __future__ import print_function
import codecs
import commands
import os
import time
import logging
import yaml
import tensorflow as tf
import numpy as np
from attrdict import AttrDict
from argparse import ArgumentParser
from tempfile import mkstemp
from tensorflow.python import debug as tf_debug
from core import utils
from models.base_model import BaseModel
from models.transformer_model import TransformerModel
from dataloader.base_dataloader import DataLoader
from models.bd_transformer_model import BD_TransformerModel
is_debug = False #False
class Evaluator(object):
"""
Evaluate the model.
"""
def __init__(self, config):
self.config = config
def init_from_config(self, config):
self.model = eval(config.model)(config, config.test.num_gpus)
self.model.build_test_model()
sess_config = tf.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
self.sess = tf.Session(config=sess_config, graph=self.model.graph)
if is_debug:
self.sess = tf_debug.LocalCLIDebugWrapperSession(self.sess)
# Restore model.
if config.test.checkpoint:
logging.info('Reload model in %s.' % os.path.join(
config.model_dir,config.test.checkpoint))
self.model.saver.restore(self.sess, os.path.join(
config.model_dir,config.test.checkpoint))
else:
logging.info('Reload model in %s.' % config.model_dir)
self.model.saver.restore(self.sess,
tf.train.latest_checkpoint(config.model_dir))
self.data_reader = DataLoader(config)
self.config = config
def init_from_existed(self, config, model, sess):
assert model.graph == sess.graph
self.config = config
self.sess = sess
self.model = model
self.data_reader = DataLoader(self.config)
def beam_search(self, X):
if 'BD' in self.config.model:
return self.sess.run([self.model.prediction, self.model.scores, self.model.alive_probs, self.model.finished_flags],
feed_dict=self.data_reader.expand_feed_dict(
{self.model.src_pls: X}))
else:
return self.sess.run(self.model.prediction,
feed_dict=self.data_reader.expand_feed_dict({self.model.src_pls: X}))
def loss(self, X, Y):
return self.sess.run(self.model.loss_sum,
feed_dict=self.data_reader.expand_feed_dict(
{self.model.src_pls: X, self.model.dst_pls: Y}))
def post_process(self, Y):
new_Y = []
for line in Y:
# 4 represents <r2l>, reverse the predicted target
if line[0] == 4:
length = np.where(line==3)
if len(length[0]) == 0:
new_line = line.tolist()[1:][::-1]
logging.info('nothing decoded--<r2l>')
else:
line = line.tolist()
new_line = line[1:length[0][0]][::-1] + line[length[0][0]:]
new_Y.append(new_line)
else:
new_line = line.tolist()
new_Y.append(new_line[1:])
return np.array(new_Y)
def translate(self, src_path, output_path):
logging.info('Translate %s.' % src_path)
tmp = output_path + '.tmp'
decode_dir = ('/').join(tmp.split('/')[:-1])
if not os.path.exists(decode_dir):
os.makedirs(decode_dir)
fd = codecs.open(tmp, 'w', 'utf8')
count = 0
token_count = 0
start = time.time()
batch_size = self.config.test.batch_size * self.config.test.num_gpus
for X,uttids in self.data_reader.get_test_batches(src_path, batch_size):
# if bd, post process
if 'BD' in self.config.model:
# print scores for debug
Y, scores, alive_probs, finished_flags = self.beam_search(X)
Y = self.post_process(Y)
else:
Y = self.beam_search(X)
sents = self.data_reader.indices_to_words(Y)
assert len(X) == len(sents)
for sent, uttid in zip(sents, uttids):
print(uttid + '\t' + sent, file=fd)
count += len(X)
token_count += np.sum(np.not_equal(Y, 3))
if token_count == 0:
print(Y.shape)
continue
time_span = time.time() - start
logging.info('{0} sentences ({1} tokens) processed in {2:.2f} minutes (speed: {3:.4f} sec/token).'.
format(count, token_count, time_span / 60, time_span / token_count))
fd.close()
# Remove BPE flag, if have.
os.system("sed -r 's/(@@ )|(@@ ?$)//g' %s > %s" % (tmp, output_path))
os.remove(tmp)
logging.info('The result file was saved in %s.' % output_path)
def ppl(self, src_path, dst_path, batch_size):
logging.info('Calculate PPL for %s and %s.' % (src_path, dst_path))
token_count = 0
loss_sum = 0
for batch in self.data_reader.get_test_batches_with_target(
src_path, dst_path, batch_size):
X, Y = batch
loss_sum += self.loss(X, Y)
token_count += np.sum(np.greater(Y, 0))
# Compute PPL
ppl = np.exp(loss_sum / token_count)
logging.info('PPL: %.4f' % ppl)
return ppl
def evaluate(self, batch_size, **kargs):
"""Evaluate the model on dev set."""
src_path = kargs['feat_file_pattern']
output_path = kargs['output_path']
cmd = kargs['cmd'] if 'cmd' in kargs else\
"perl multi-bleu.perl {ref} < {output} 2>/dev/null | awk '{{print($3)}}' | awk -F, '{{print $1}}'"
self.translate(src_path, output_path)
# if 'ref_path' in kargs:
# ref_path = kargs['ref_path']
# bleu = commands.getoutput(cmd.format(**{'ref': ref_path, 'output': output_path}))
# logging.info('BLEU: {}'.format(bleu))
# return float(bleu)
# if 'dst_path' in kargs:
# self.ppl(src_path, kargs['dst_path'], batch_size)
return None
if __name__ == '__main__':
from ctypes import cdll
#cdll.LoadLibrary('/usr/local/cuda/lib64/libcudnn.so')
cdll.LoadLibrary('/usr/local/cuda-9.0/lib64/libcudnn.so')
import os
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
#os.environ["CUDA_VISIBLE_DEVICES"] = "1"
# os.environ["CUDA_VISIBLE_DEVICES"] = "3"
parser = ArgumentParser()
parser.add_argument('-c', '--config', dest='config')
parser.add_argument('-ch', '--avg_checkpoint', dest='avg_checkpoint')
args = parser.parse_args()
# Read config
if not args.config:
args.config = './config_template_pinyin.yaml'
config = yaml.load(open(args.config))
if args.avg_checkpoint:
config['test']['checkpoint'] = 'averaged.ckpt-0'
config['test']['set1']['output_path'] = '/'.join(config['test']['set1']['output_path'].split('/')[:-1])+'/averaged_result.txt'
config = AttrDict(config)
# Logger
logging.basicConfig(level=logging.INFO)
evaluator = Evaluator(config)
evaluator.init_from_config(config)
for attr in config.test:
if attr.startswith('set'):
evaluator.evaluate(config.test.batch_size * config.test.num_gpus,
**config.test[attr])
logging.info("Done")