tensorflow环境配置记录备忘

eumnq8

pip install tensorflow==1.15  -i  https://pypi.tuna.tsinghua.edu.cn/simple  --default-timeout=1000

pip install  tensorflow-serving-api==1.15  -i  https://pypi.tuna.tsinghua.edu.cn/simple  --default-timeout=1000

ubuntu16.04安装docker步骤：（必须安装特定版本）
1、
sudo apt-get remove docker docker-engine docker.io
2、
sudo apt-get update
3、
sudo apt-get install \
      apt-transport-https \
      ca-certificates \
      curl \
      software-properties-common

4、
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -

5、
sudo add-apt-repository \
     "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
     $(lsb_release -cs) \
     stable"

6、
sudo apt-get update

7、
sudo apt-get install docker-ce=17.03.2~ce-0~ubuntu-xenial

sudo service docker start
sudo docker version

ubuntu16.04更新阿里源：
https://www.cnblogs.com/mufire/p/6433757.html

https://blog.csdn.net/u012433049/article/details/88972850

部署
git clone https://github.com/tensorflow/serving

设置docker镜像
curl -sSL https://get.daocloud.io/daotools/set_mirror.sh | sh -s http://f1361db2.m.daocloud.io
sudo service docker restart,重启docker，然后拉取tf_serve

sudo docker images  查看镜像有没有

用下面那个
sudo docker run --runtime=cpu -p 8501:8501 \
        --mount type=bind,\
        source=/home/szl/serving/tensorflow_serving/servables/tensorflow/testdata/saved_model_half_plus_two_cpu,\
        target=/models/half_plus_two \
        -e MODEL_NAME=half_plus_two -t tensorflow/serving &

sudo docker run -t --rm -p 8501:8501 \
    -v "$(pwd)/serving/tensorflow_serving/servables/tensorflow/testdata/saved_model_half_plus_two_cpu:/models/half_plus_two" \
    -e MODEL_NAME=half_plus_two \
    tensorflow/serving &

sudo curl -d '{"instances": [1.0, 2.0, 5.0]}' \
-X POST http://localhost:8501/v1/models/half_plus_two:predict

查看容器id
sudo docker ps

pb文件生成：ddz_saved_model.py
使用demo：play_game_client.py

ckpt是已经训练好的checkpoint模型文件