-
Notifications
You must be signed in to change notification settings - Fork 0
/
main1.py
357 lines (295 loc) · 13.6 KB
/
main1.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
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
import os
import subprocess
os.environ['PYTHONIOENCODING'] = 'utf-8'
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1'
import json
import numpy as np
import random
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn.utils.rnn import pack_padded_sequence
from pytorch_transformers.tokenization_bert import BertTokenizer
from pytorch_transformers.modeling_bert import BertConfig, BertModel
from pytorch_transformers import AdamW
from logger import get_logger
from parser import get_parser
seed = 77
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
parser = get_parser()
option = parser.parse_args()
root_path = 'result'
logs_folder = os.path.join(root_path, 'logs', option.name)
save_folder = os.path.join(root_path, 'save', option.name)
sample_folder = os.path.join(root_path, 'sample', option.name)
result_folder = os.path.join(root_path, 'result', option.name)
logs_path = option.name + '.log'
save_path = option.name + '.bin'
sample_path = option.name + '.csv'
result_path = option.name + '.csv'
subprocess.run('mkdir -p %s' % logs_folder, shell = True)
subprocess.run('mkdir -p %s' % save_folder, shell = True)
subprocess.run('mkdir -p %s' % sample_folder, shell = True)
subprocess.run('mkdir -p %s' % result_folder, shell = True)
logger = get_logger(option.name, os.path.join(logs_folder, logs_path))
logger.info('Prepare Data')
batch_size = option.batch_size
max_seq_len = option.max_seq_len
train_input = option.train_input
valid_input = option.valid_input
test_input = option.test_input
model_name_or_path = option.model_name_or_path
trained_model_path = os.path.join(save_folder, save_path)
sample_file = os.path.join(sample_folder, sample_path)
result_file = os.path.join(result_folder, result_path)
tokenizer = BertTokenizer.from_pretrained(model_name_or_path, do_lower_case = True)
class InputExample():
def __init__(self, guid, source, target):
self.guid = guid
self.source = source
self.target = target
class InputFeature():
def __init__(self, guid, features, labels):
self.guid = guid
tokens, input_ids, input_mask, segment_ids = features
self.features = {
'tokens': tokens,
'input_ids': input_ids,
'input_mask': input_mask,
'segment_ids': segment_ids,
'labels': labels
}
def read_examples(file_path):
examples = []
with open(file_path, 'rt', encoding = 'utf-8') as jsonl_file:
for index, line in enumerate(jsonl_file):
data = json.loads(line)
text = data['content']
idioms = data['groundTruth']
for idiom in idioms:
text = text.replace('#idiom#', idiom, 1)
source = list(text)
target = ['O'] * len(text)
for idiom in idioms:
start_idx = text.index(idiom)
target[start_idx: start_idx + len(idiom)] = ['I'] * len(idiom)
example = InputExample(index, source, target)
examples.append(example)
return examples
def convert_examples_to_features(examples, tokenizer, max_seq_len):
features = []
mapping = {'O': 0, 'I': 1, 'B': 2}
for example in examples:
tokens = []
labels = []
source = example.source
target = example.target
for s, t in zip(source, target):
splits = tokenizer.tokenize(s)
if len(splits) > 0:
tokens.extend(splits)
labels.extend([t] + ['O'] * (len(splits) - 1))
tokens = tokens[:max_seq_len - 1]
labels = labels[:max_seq_len - 1]
tokens.insert(0, '[CLS]')
labels.insert(0, 'O')
input_ids = tokenizer.convert_tokens_to_ids(tokens)
input_mask = [1] * len(tokens)
segment_ids = [0] * len(tokens)
label_ids = [mapping[label] for label in labels]
while len(input_ids) < max_seq_len:
tokens.append('[PAD]')
input_ids.append(0)
input_mask.append(0)
segment_ids.append(0)
label_ids.append(0)
feature = InputFeature(
example.guid,
(tokens, input_ids, input_mask, segment_ids),
label_ids
)
features.append(feature)
return features
def select_field(features, field):
return [
feature.features[field] for feature in features
]
train_examples = read_examples(train_input)
valid_examples = read_examples(valid_input)
test_examples = read_examples(test_input )
train_features = convert_examples_to_features(train_examples, tokenizer, max_seq_len)
valid_features = convert_examples_to_features(valid_examples, tokenizer, max_seq_len)
test_features = convert_examples_to_features(test_examples , tokenizer, max_seq_len)
train_tokens = select_field(train_features, 'tokens')
train_input_ids = np.array(select_field(train_features, 'input_ids'))
train_input_mask = np.array(select_field(train_features, 'input_mask'))
train_segment_ids = np.array(select_field(train_features, 'segment_ids'))
train_labels = np.array(select_field(train_features, 'labels'))
valid_tokens = select_field(valid_features, 'tokens')
valid_input_ids = np.array(select_field(valid_features, 'input_ids'))
valid_input_mask = np.array(select_field(valid_features, 'input_mask'))
valid_segment_ids = np.array(select_field(valid_features, 'segment_ids'))
valid_labels = np.array(select_field(valid_features, 'labels'))
test_tokens = select_field(test_features, 'tokens')
test_input_ids = np.array(select_field(test_features, 'input_ids'))
test_input_mask = np.array(select_field(test_features, 'input_mask'))
test_segment_ids = np.array(select_field(test_features, 'segment_ids'))
test_labels = np.array(select_field(test_features, 'labels'))
train_input_ids_tensor = torch.tensor(train_input_ids, dtype = torch.long)
train_input_mask_tensor = torch.tensor(train_input_mask, dtype = torch.long)
train_segment_ids_tensor = torch.tensor(train_segment_ids, dtype = torch.long)
train_label_tensor = torch.tensor(train_labels, dtype = torch.long)
valid_input_ids_tensor = torch.tensor(valid_input_ids, dtype = torch.long)
valid_input_mask_tensor = torch.tensor(valid_input_mask, dtype = torch.long)
valid_segment_ids_tensor = torch.tensor(valid_segment_ids, dtype = torch.long)
valid_label_tensor = torch.tensor(valid_labels, dtype = torch.long)
test_input_ids_tensor = torch.tensor(test_input_ids, dtype = torch.long)
test_input_mask_tensor = torch.tensor(test_input_mask, dtype = torch.long)
test_segment_ids_tensor = torch.tensor(test_segment_ids, dtype = torch.long)
test_label_tensor = torch.tensor(test_labels, dtype = torch.long)
train_dataset = torch.utils.data.TensorDataset(train_input_ids_tensor, train_input_mask_tensor, train_segment_ids_tensor, train_label_tensor)
valid_dataset = torch.utils.data.TensorDataset(valid_input_ids_tensor, valid_input_mask_tensor, valid_segment_ids_tensor, valid_label_tensor)
test_dataset = torch.utils.data.TensorDataset(test_input_ids_tensor , test_input_mask_tensor , test_segment_ids_tensor , test_label_tensor )
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size = batch_size, shuffle = True )
valid_loader = torch.utils.data.DataLoader(valid_dataset, batch_size = batch_size, shuffle = False)
test_loader = torch.utils.data.DataLoader(test_dataset , batch_size = batch_size, shuffle = False)
logger.info('Prepare Model')
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
hidden_size = option.hidden_size
num_classes = option.num_classes
class BertForTokenClassification(nn.Module):
def __init__(self, model_name_or_path, hidden_size = 768, num_classes = 2):
super(BertForTokenClassification, self).__init__()
self.config = BertConfig.from_pretrained(model_name_or_path)
self.bert = BertModel.from_pretrained(model_name_or_path, config = self.config)
for param in self.bert.parameters():
param.requires_grad = True
self.dropout = nn.Dropout(0.5)
self.linear = nn.Linear(hidden_size, num_classes)
def forward(self, input_ids, input_mask, segment_ids):
last_hidden_states, _ = self.bert(input_ids, attention_mask = input_mask, token_type_ids = segment_ids)
output = self.linear(self.dropout(last_hidden_states))
return output
model = BertForTokenClassification(model_name_or_path, hidden_size, num_classes)
model = nn.DataParallel(model)
model = model.to(device)
logger.info('Train & Valid')
n_epoch = option.n_epoch
learning_rate = option.learning_rate
parameters = list(model.named_parameters())
no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
grouped_parameters = [
{'params': [p for n, p in parameters if not any(nd in n for nd in no_decay)], 'weight_decay': 0.01},
{'params': [p for n, p in parameters if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}
]
criterion = torch.nn.CrossEntropyLoss()
optimizer = AdamW(grouped_parameters, lr = learning_rate, eps = 1e-6)
def extract(token, pred):
pred_true = np.where(pred == 1)[0]
pred_diff = np.diff (pred_true)
seg_point = np.where(pred_diff != 1)[0] + 1
seg_group = np.split(list(pred_true), seg_point)
token = np.array(token)
idioms = []
while len(seg_group) > 0:
group = seg_group.pop()
if len(group) == 4:
idom = token[group]
idioms.append(idom)
elif len(group) > 4:
seg_group.append(group[:4])
seg_group.append(group[4:])
elif len(group) < 4:
continue
idioms.reverse()
idioms = [''.join(idiom) for idiom in idioms]
return idioms
def score(tokens, y_true, y_pred):
A, B, C = 1e-10, 1e-10, 1e-10
for i in range(len(tokens)):
R = set(extract(tokens[i], y_pred[i]))
T = set(extract(tokens[i], y_true[i]))
A += len(R & T)
B += len(R)
C += len(T)
precision, recall, f1 = A / B, A / C, 2 * A / (B + C)
return f1, precision, recall
def save2file(tokens, y_true, y_pred, file_path):
texts = []
true_idioms = []
pred_idioms = []
for i in range(len(tokens)):
texts.append(''.join(tokens[i]).lstrip('[CLS]').rstrip('[PAD]*'))
true_idioms.append(';'.join(extract(tokens[i], y_true[i])))
pred_idioms.append(';'.join(extract(tokens[i], y_pred[i])))
with open(file_path, 'w', encoding = 'utf-8') as csv_file:
csv_file.write('"text","true_idiom","pred_idiom"\n')
for text, true_idiom, pred_idiom in zip(texts, true_idioms, pred_idioms):
csv_file.write(
'"' + text + '"' + ',' +
'"' + true_idiom + '"' + ',' +
'"' + pred_idiom + '"' + '\n'
)
best_f1 = 0
patience = 5
early_stop = 0
for epoch in range(1, n_epoch + 1):
model.train()
train_loss = 0
for _, batch in enumerate(train_loader):
batch = [data.to(device) for data in batch]
x_ids, x_mask, x_seg_ids, y_true = batch
y_pred = model(x_ids, x_mask, x_seg_ids)
length = torch.sum(x_mask, dim = 1).cpu()
y_pred_pad = pack_padded_sequence(y_pred, length, batch_first = True, enforce_sorted = False).data
y_true_pad = pack_padded_sequence(y_true, length, batch_first = True, enforce_sorted = False).data
loss = criterion(y_pred_pad, y_true_pad)
train_loss += loss.item()
optimizer.zero_grad()
loss.backward()
optimizer.step()
model.eval()
valid_loss = 0
valid_pred_fold = np.zeros((valid_labels.shape[0], valid_labels.shape[1], num_classes))
with torch.no_grad():
for i, batch in enumerate(valid_loader):
batch = [data.to(device) for data in batch]
x_ids, x_mask, x_seg_ids, y_true = batch
y_pred = model(x_ids, x_mask, x_seg_ids)
length = torch.sum(x_mask, dim = 1).cpu()
y_pred_pad = pack_padded_sequence(y_pred, length, batch_first = True, enforce_sorted = False).data
y_true_pad = pack_padded_sequence(y_true, length, batch_first = True, enforce_sorted = False).data
loss = criterion(y_pred_pad, y_true_pad)
valid_loss += loss.item()
valid_pred_fold[i * batch_size: (i + 1) * batch_size] = F.softmax(y_pred, dim = 2).detach().cpu().numpy()
f1, precision, recall = score(valid_tokens, valid_labels, np.argmax(valid_pred_fold, axis = 2))
if best_f1 < f1:
best_f1 = f1
early_stop = 0
torch.save(model.state_dict(), trained_model_path)
save2file(valid_tokens, valid_labels, np.argmax(valid_pred_fold, axis = 2), sample_file)
else:
early_stop += 1
logger.info(
'epoch: %d, train_loss: %.8f, valid_loss: %.8f, precision: %.8f, recall: %.8f, f1: %.8f, best_f1: %.8f' %
(epoch, train_loss / len(train_loader), valid_loss / len(valid_loader), precision, recall, f1, best_f1)
)
torch.cuda.empty_cache()
if early_stop > patience:
break
model.load_state_dict(torch.load(trained_model_path))
model.eval()
test_preds_fold = np.zeros((test_labels.shape[0], test_labels.shape[1], num_classes))
with torch.no_grad():
for i, batch in enumerate(test_loader):
batch = [data.to(device) for data in batch]
x_ids, x_mask, x_seg_ids, y_true = batch
y_pred = model(x_ids, x_mask, x_seg_ids)
test_preds_fold[i * batch_size: (i + 1) * batch_size] = F.softmax(y_pred, dim = 2).detach().cpu().numpy()
f1, precision, recall = score(test_tokens, test_labels, np.argmax(test_preds_fold, axis = 2))
logger.info('epoch: best, precision: %.8f, recall: %.8f, f1: %.8f' % (precision, recall, f1))
save2file(test_tokens, test_labels, np.argmax(test_preds_fold, axis = 2), result_file)