pip3 install --user --upgrade tensorflow 

安装遇到的问题如下：

pip3 install --user --upgrade tensorflow 
 1052  pip uninstall protobuf
 1053  pip3 uninstall protobuf
 1054  pip3 install protobuf==3.20.*
 1055  pip3 install open-clip-torch==2.8.2
 1056  pip3 install --user --upgrade tensorflow 
 1057  pip uninstall protobuf
 1058  pip3 uninstall protobuf
 1059  pip3 install protobuf==3.20.*
 1060  pip3 install --user --upgrade tensorflow 
 1061  pip install protobuf==3.20.*
 

tensorflow-2.11

bert-as-service, 依赖于 python≥3.5 AND tensorflow≥1.10;

参考：Bert- as-service安装和使用教程_bert进入端口-CSDN博客

参考：https://tensorflow.google.cn/install/pip?hl=zh-cn#system-install

BERT 服务化 bert-as-service - 01码匠 - 博客园 (cnblogs.com)

https://blog.csdn.net/luhuibo318/article/details/104925892

下载 https://storage.googleapis.com/bert_models/2018_11_03/chinese_L-12_H-768_A-12.zip

pip3 install bert-serving-server

pip3 install bert-serving-client

pip install bert-serving-server

pip install bert-serving-client

1. 进入chinese_L-12_H-768_A-12同级目录执行下面命令

python3 start-bert-as-service.py -model_dir /opt/dockerinstall/python3/bert-model/chinese_L-12_H-768_A-12/ -num_worker=4

bert-serving-start -model_dir /opt/dockerinstall/python3/bert-model/chinese_L-12_H-768_A-12/ -num_worker=4

import sys

from bert_serving.server import BertServer
from bert_serving.server.helper import get_run_args

if __name__ == '__main__':
    args = get_run_args()
    server = BertServer(args)
    server.start()
    server.join()

tensorflow 版本太高

安装 pip3 install tensorflow==1.14.0

sudo yum update libstdc++
参考：https://blog.csdn.net/my__blog/article/details/123703024

https://blog.csdn.net/qq_50665031/article/details/108987205

最后启动成功：

待测试：nohup bert-serving-start -model_dir ./chinese_L-12_H-768_A-12/ -num_worker=1 > bert.log 2>&1 &