-
Notifications
You must be signed in to change notification settings - Fork 0
/
squad_qg.py
186 lines (161 loc) · 6.35 KB
/
squad_qg.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
#! -*- coding: utf-8 -*-
# WoBERT做Seq2Seq任务，采用UniLM方案
# 介绍链接：https://kexue.fm/archives/6933
# 数据集：https://github.com/CLUEbenchmark/CLGE 中的LCSTS数据集
# 补充了评测指标bleu、rouge-1、rouge-2、rouge-l
from __future__ import print_function
import json
import numpy as np
from tqdm import tqdm
from bert4keras.backend import keras, K
from bert4keras.layers import Loss
from bert4keras.models import build_transformer_model
from bert4keras.tokenizers import Tokenizer, load_vocab
from bert4keras.optimizers import Adam
from bert4keras.snippets import sequence_padding, open
from bert4keras.snippets import DataGenerator, AutoRegressiveDecoder
from keras.models import Model
from rouge import Rouge # pip install rouge
from nltk.translate.bleu_score import sentence_bleu, SmoothingFunction
import jieba
jieba.initialize()
# 基本参数
maxlen = 512
batch_size = 16
epochs = 40
# bert配置
config_path = 'google_bert/bert_config.json'
checkpoint_path = 'google_bert/bert_model.ckpt'
dict_path = 'google_bert/vocab.txt'
# def load_data(filename):
# D = []
# with open(filename, encoding='utf-8') as f:
# data = json.load(f)
# for l in data:
# title, content = l['tgt'], l['src']
# D.append((title, content[:256]))
# return D
# 加载数据集
# train_data = load_data('data/tmp_data/train.json')
# valid_data = load_data('data/tmp_data/dev.json')
# test_data = load_data('data/tmp_data/test.json')
def load_data(file_name):
D = []
import os
with open(os.path.join(file_name, 'train.pa.10000.txt'), encoding='utf-8') as f1, \
open(os.path.join(file_name, 'train.q.10000.txt'), encoding='utf-8') as f2:
for pa, q in zip(f1, f2):
D.append([q, pa])
train_data, dev_data = D[: int(len(D)*0.7)], D[int(len(D)*0.7):]
return train_data, dev_data
train_data, valid_data = load_data('squad_data')
test_data = valid_data
# 建立分词器
tokenizer = Tokenizer(
dict_path,
do_lower_case=True,
# pre_tokenize=lambda s: jieba.cut(s, HMM=False)
)
class data_generator(DataGenerator):
"""数据生成器
"""
def __iter__(self, random=False):
batch_token_ids, batch_segment_ids = [], []
for is_end, (title, content) in self.sample(random):
token_ids, segment_ids = tokenizer.encode(
content, title, maxlen=maxlen
)
batch_token_ids.append(token_ids)
batch_segment_ids.append(segment_ids)
if len(batch_token_ids) == self.batch_size or is_end:
batch_token_ids = sequence_padding(batch_token_ids)
batch_segment_ids = sequence_padding(batch_segment_ids)
yield [batch_token_ids, batch_segment_ids], None
batch_token_ids, batch_segment_ids = [], []
class CrossEntropy(Loss):
"""交叉熵作为loss，并mask掉输入部分
"""
def compute_loss(self, inputs, mask=None):
y_true, y_mask, y_pred = inputs
y_true = y_true[:, 1:] # 目标token_ids
y_mask = y_mask[:, 1:] # segment_ids，刚好指示了要预测的部分
y_pred = y_pred[:, :-1] # 预测序列，错开一位
loss = K.sparse_categorical_crossentropy(y_true, y_pred)
loss = K.sum(loss * y_mask) / K.sum(y_mask)
return loss
model = build_transformer_model(
config_path, checkpoint_path, application='unilm'
)
output = CrossEntropy(2)(model.inputs + model.outputs)
model = Model(model.inputs, output)
model.compile(optimizer=Adam(1e-5))
model.summary()
class AutoTitle(AutoRegressiveDecoder):
"""seq2seq解码器
"""
@AutoRegressiveDecoder.wraps(default_rtype='probas')
def predict(self, inputs, output_ids, states):
token_ids, segment_ids = inputs
token_ids = np.concatenate([token_ids, output_ids], 1)
segment_ids = np.concatenate([segment_ids, np.ones_like(output_ids)], 1)
return model.predict([token_ids, segment_ids])[:, -1]
def generate(self, text, topk=1):
max_c_len = maxlen - self.maxlen
token_ids, segment_ids = tokenizer.encode(text, maxlen=max_c_len)
# token_ids = [token_ids, token_ids]
# segment_ids = [segment_ids, segment_ids]
output_ids = self.beam_search([token_ids, segment_ids],
topk) # 基于beam search
return tokenizer.decode(output_ids)
autotitle = AutoTitle(start_id=None, end_id=tokenizer._token_end_id, maxlen=32)
class Evaluator(keras.callbacks.Callback):
def __init__(self):
self.rouge = Rouge()
self.smooth = SmoothingFunction().method1
self.best_bleu = 0.
def on_epoch_end(self, epoch, logs=None):
metrics = self.evaluate(valid_data) # 评测模型
if metrics['bleu'] > self.best_bleu:
self.best_bleu = metrics['bleu']
model.save_weights('./best_model_lcsts.weights') # 保存模型
metrics['best_bleu'] = self.best_bleu
print('valid_data:', metrics)
def evaluate(self, data, topk=1):
total = 0
rouge_1, rouge_2, rouge_l, bleu = 0, 0, 0, 0
for title, content in tqdm(data):
total += 1
title = ' '.join(title).lower()
pred_title = ' '.join(autotitle.generate(content, topk)).lower()
if pred_title.strip():
scores = self.rouge.get_scores(hyps=pred_title, refs=title)
rouge_1 += scores[0]['rouge-1']['f']
rouge_2 += scores[0]['rouge-2']['f']
rouge_l += scores[0]['rouge-l']['f']
bleu += sentence_bleu(
references=[title.split(' ')],
hypothesis=pred_title.split(' '),
smoothing_function=self.smooth
)
rouge_1 /= total
rouge_2 /= total
rouge_l /= total
bleu /= total
return {
'rouge-1': rouge_1,
'rouge-2': rouge_2,
'rouge-l': rouge_l,
'bleu': bleu,
}
if __name__ == '__main__':
evaluator = Evaluator()
train_generator = data_generator(train_data, batch_size)
print(autotitle.generate("hello, my name is hetongxue, nice to meet you"))
model.fit_generator(
train_generator.forfit(),
steps_per_epoch=len(train_generator),
epochs=epochs,
callbacks=[evaluator]
)
else:
model.load_weights('./best_model_lcsts.weights')