bert-component/component_bert_sequence_label_predict.py at master · YC-wind/bert-component · GitHub

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
#!/usr/bin/python
# coding:utf8
"""
@author: Cong Yu
@time: 2019-06-06 15:18
"""
import tensorflow as tf
from sklearn.externals import joblib
import numpy as np
import pandas as pd
import json, time

# model_folder = "./bin"

labels = ["<pad>", "[CLS]", "[SEP]", "O", "B-PER", "I-PER", "B-ORG", "I-ORG", "B-LOC", "I-LOC"]

config = {
    "in_1": "./out/tokenizer.m",  # 第一个输入为 tokenizer 序列化模型(由上一次传递过来)
    "in_2": "./bin_seq/",
    "file": "./seq_dev.csv",  # 第二个输入为 训练/测试 文件
    "column_name_x1": "text",
    "label_list": labels,  # 整个样本空间的 标签集
    "split": "",  # 标签的分割符，默认为空，表示单标签，不为空的化，按分隔符进行分割出多标签
    "max_seq_len": 128,  # 输入文本片段的最大 char级别 长度
    "out": "./data/predict_seq_dev.csv"  # 输出为 tf_record 的二进制文件
}


def process_one_example(tokenizer, text, max_seq_len=128):
    textlist = text.split(' ')
    tokens = []
    labels = []
    for i, word in enumerate(textlist):
        token = tokenizer.tokenize(word)
        # print(token)
        tokens.extend(token)
        # label_1 = labellist[i]
        # print(label_1) 可能会拆成多个
        # for m in range(len(token)):
        #     if m == 0:
        #         labels.append(label_1)
        #     else:
        #         labels.append(labels[0])
    # tokens = tokenizer.tokenize(example.text)  -2 的原因是因为序列需要加一个句首和句尾标志
    if len(tokens) >= max_seq_len - 1:
        tokens = tokens[0:(max_seq_len - 2)]
        labels = labels[0:(max_seq_len - 2)]
    ntokens = []
    segment_ids = []
    label_ids = []
    ntokens.append("[CLS]")  # 句子开始设置CLS 标志
    segment_ids.append(0)
    # append("O") or append("[CLS]") not sure!
    # label_ids.append(label2id["[CLS]"])
    for i, token in enumerate(tokens):
        ntokens.append(token)
        segment_ids.append(0)
        # label_ids.append(label2id[labels[i]])
    ntokens.append("[SEP]")
    segment_ids.append(0)
    # append("O") or append("[SEP]") not sure!
    # label_ids.append(label2id["[SEP]"])
    input_ids = tokenizer.convert_tokens_to_ids(ntokens)
    input_mask = [1] * len(input_ids)
    # label_mask = [1] * len(input_ids)
    while len(input_ids) < max_seq_len:
        input_ids.append(0)
        input_mask.append(0)
        segment_ids.append(0)
        # we don't concerned about it!
        label_ids.append(0)
        ntokens.append("**NULL**")
        # label_mask.append(0)
    # print(len(input_ids))
    assert len(input_ids) == max_seq_len
    assert len(input_mask) == max_seq_len
    assert len(segment_ids) == max_seq_len
    # assert len(label_ids) == max_seq_len

    feature = (input_ids, input_mask, segment_ids)
    return feature


def load_model(model_folder):
    # We retrieve our checkpoint fullpath
    try:
        checkpoint = tf.train.get_checkpoint_state(model_folder)
        input_checkpoint = checkpoint.model_checkpoint_path
        print("[INFO] input_checkpoint:", input_checkpoint)
    except Exception as e:
        input_checkpoint = model_folder
        print("[INFO] Model folder", model_folder, repr(e))

    # We clear devices to allow TensorFlow to control on which device it will load operations
    clear_devices = True

    # We import the meta graph and retrieve a Saver
    saver = tf.train.import_meta_graph(input_checkpoint + '.meta', clear_devices=clear_devices)

    # We retrieve the protobuf graph definition
    graph = tf.get_default_graph()
    input_graph_def = graph.as_graph_def()

    # We start a session and restore the graph weights
    sess = tf.Session()
    saver.restore(sess, input_checkpoint)
    return sess


def main():
    tokenizer = joblib.load(config["in_1"])
    sess = load_model(config["in_2"])

    input_ids = sess.graph.get_tensor_by_name("input_ids:0")
    input_mask = sess.graph.get_tensor_by_name("input_mask:0")  # is_training
    segment_ids = sess.graph.get_tensor_by_name("segment_ids:0")  # fc/dense/Relu  cnn_block/Reshape
    keep_prob = sess.graph.get_tensor_by_name("keep_prob:0")

    p = sess.graph.get_tensor_by_name("loss/tag:0")

    df = pd.read_csv(config["file"], index_col=0)
    questions = []
    predicts = []
    count = 0
    t1 = time.time()
    for index, row in df.iterrows():
        # label = label2id[row["topic"].strip()]
        if not (row[config["column_name_x1"]]):
            continue
        if not isinstance(row[config["column_name_x1"]], str):
            print(row[config["column_name_x1"]])
            continue
        feature = process_one_example(tokenizer, row[config["column_name_x1"]], max_seq_len=config["max_seq_len"])
        if count < 5:
            print(feature[0])
            print(feature[1])
            print(feature[2])

        questions.append(row[config["column_name_x1"]])
        feed = {input_ids: [feature[0]],
                input_mask: [feature[1]],
                segment_ids: [feature[2]],
                keep_prob: 1.0
                }

        probs = sess.run([p], feed)[0][0]
        result = []

        len_ = len(row[config["column_name_x1"]].split(" "))
        for ii, v in enumerate(probs[1:len_ + 1]):
            result.append(config["label_list"][int(v)])

        predicts.append(json.dumps(result, ensure_ascii=False))
        count += 1
        if count == 100:
            break
    t2 = time.time()
    print("predict cost time:", t2 - t1)
    df_out = pd.DataFrame()
    df_out["question"] = questions
    df_out["predict"] = predicts
    df_out.to_csv(config["out"])


if __name__ == "__main__":
    print("********* component_bert_sequence_label_predict start *********")
    main()