bert模型docker部署

bert模型docker部署 1. bert模型文件ckpt转化为pb文件

在run_classifier.py文件中添加函数serving_input_fn
def serving_input_fn():
    # 保存模型为SaveModel格式
    # 采用最原始的feature方式，输入是feature Tensors。
    # 如果采用build_parsing_serving_input_receiver_fn，则输入是tf.Examples
    label_ids = tf.placeholder(tf.int32, [None], name='label_ids')  # 要素识别任务有2个类别
    input_ids = tf.placeholder(tf.int32, [None, FLAGS.max_seq_length], name='input_ids')
    input_mask = tf.placeholder(tf.int32, [None, FLAGS.max_seq_length], name='input_mask')
    segment_ids = tf.placeholder(tf.int32, [None, FLAGS.max_seq_length], name='segment_ids')
    input_fn = tf.estimator.export.build_raw_serving_input_receiver_fn({
        'label_ids': label_ids,
        'input_ids': input_ids,
        'input_mask': input_mask,
        'segment_ids': segment_ids,
    })()
    return input_fn
    
在main函数中修改验证代码
    if FLAGS.do_eval:
        estimator._export_to_tpu = False
        estimator.export_savedmodel(FLAGS.trans_model_dir, serving_input_fn)

2. docker部署bert模型

• 查看centos的版本

cat /etc/redhat-release

• 安装docker

yum install docker

• 启动docker

service docker start

• 查看docker安装的镜像

docker images

• 安装tensorflow-serving镜像

docker pull tensorflow/serving

• 在docker上启动镜像生成容器

docker run -t --rm -p 5000:8501 -v /opt/fuzzy_model/output:/models/bert-model -e MODEL_NAME=bert-model tensorflow/serving &

目录结构如下：

3. 使用docker容器接口

# -*- coding: utf-8 -*-
"""
Created on Fri Mar  6 14:41:13 2020
@author: 
"""
import json
import os
import requests
from fuzzy_query import tokenization
import numpy as np


def get_work_dir():
    return os.path.dirname(
        os.path.dirname(
            os.path.dirname(
                os.path.dirname(__file__)
            )
        )
    )

vocal_file_path = 'vocab.txt'
# vocal_file_path = os.path.join(get_work_dir(), 'var', 'vocab.txt')
tokenizer = tokenization.FullTokenizer(vocab_file=vocal_file_path,
                                       do_lower_case=True)
max_seq_length = 64


def text2ids(text_list):
    input_ids_list = []
    input_mask_list = []
    label_ids_list = []
    segment_ids_list = []
    for text in text_list:
        label = 0
        tokens_a = tokenizer.tokenize(text)
        if len(tokens_a) > max_seq_length - 2:
            tokens_a = tokens_a[0:(max_seq_length - 2)]
        tokens = []
        segment_ids = []
        tokens.append("[CLS]")
        segment_ids.append(0)
        for token in tokens_a:
            tokens.append(token)
            segment_ids.append(0)
        tokens.append("[SEP]")
        segment_ids.append(0)
        input_ids = tokenizer.convert_tokens_to_ids(tokens)
        input_mask = [1] * len(input_ids)
        while len(input_ids) < max_seq_length:
            input_ids.append(0)
            input_mask.append(0)
            segment_ids.append(0)
        input_ids_list.append(input_ids)
        input_mask_list.append(input_mask)
        label_ids_list.append(label)
        segment_ids_list.append(segment_ids)
    return input_ids_list, input_mask_list, label_ids_list, segment_ids_list


def bert_class(textList):
    """
    调用tfserving服务的接口，对外提供服务
    :parma textList: 输入的文本列表
    :return result: 结果
    """
    input_ids_list, input_mask_list, label_ids_list, segment_ids_list = 
        text2ids(textList)
    data = json.dumps(
        {
            "name": 'bert',
            "signature_name": 'serving_default',
            "inputs": {
                'input_ids': input_ids_list,
                'input_mask': input_mask_list,
                'label_ids': label_ids_list,
                'segment_ids': segment_ids_list
            }})
    headers = {"content-type": "application/json"}
    url = 'http://10.30.239.205:5000/v1/models/bert-model:predict'
    json_response = requests.post(url, data=data, headers=headers)
    predictions = json.loads(json_response.text)['outputs']
    pre_list = [np.argmax(result) for result in predictions]
    print(pre_list)
    return pre_list


if __name__ == "__main__":
    name_list = ['饰品']
    bert_class(name_list)