Docker使用tensorflow serving部署mnist模型

参考：
1、https://tensorflow.google.cn/serving/serving_inception
2、https://tensorflow.google.cn/serving/serving_basic

---

主机安装tensorflow serving 参考这里
主机使用tensorflow serving部署mnist模型参考这里
Docker安装tensorflow serving 参考这里
Docker中部署Inception模型 参考这里

---

1、创建一个Docker镜像

参考：Docker安装tensorflow serving 参考这里

运行容器

docker pull registry.cn-hangzhou.aliyuncs.com/781708249/tensorflow-serving:v1 # 已经配置好的tensorflow serving 从阿里镜像拉下来
git clone --recurse-submodules https://github.com/tensorflow/serving # serving下载到主机上
docker run --name=mnist_container -it -v /home/wu/serving:/serving registry.cn-hangzhou.aliyuncs.com/781708249/tensorflow-serving:v1 /bin/bash # 使用-v 挂载到容器中

配置和构建TensorFlow服务

root@c97d8e820ced:/# cd serving/tensorflow
root@c97d8e820ced:/serving/tensorflow# ./configure
root@c97d8e820ced:/serving# cd ..
root@c97d8e820ced:/serving# bazel build -c opt tensorflow_serving/example/...

root@c97d8e820ced:/serving# bazel build -c opt tensorflow_serving/model_servers:tensorflow_model_server

在容器中导出初始模型

在正在运行的容器中，我们运行mnist_saved_model.py

root@c97d8e820ced:/serving# rm -rf /tmp/mnist_model
root@c97d8e820ced:/serving# bazel-bin/tensorflow_serving/example/mnist_saved_model /tmp/mnist_model
root@c97d8e820ced:/serving# [Ctrl-p] + [Ctrl-q]

提交镜像进行部署

$ docker commit mnist_container $USER/mnist_serving
$ docker stop mnist_container 

2、在本地Docker容器中运行

我们使用构建的镜像在本地测试服务工作流程。

# $ docker run -it $USER/mnist_serving
$ docker run -it -v /home/wu/serving:/serving $USER/mnist_serving

启动服务器

在容器中运行gRPC tensorflow_model_server

root@f07eec53fd95:/# cd serving
root@f07eec53fd95:/serving# bazel-bin/tensorflow_serving/model_servers/tensorflow_model_server --port=9000 --model_name=mnist --model_base_path=/tmp/mnist_model/ &> mnist_log &
[2] 80

查询服务器

使用mnist_client.py查询服务器。

root@f07eec53fd95:/serving# bazel-bin/tensorflow_serving/example/mnist_client --num_tests=1000 --server=localhost:9000

Extracting /tmp/train-images-idx3-ubyte.gz
Extracting /tmp/train-labels-idx1-ubyte.gz
Extracting /tmp/t10k-images-idx3-ubyte.gz
Extracting /tmp/t10k-labels-idx1-ubyte.gz
........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
Inference error rate: 10.4%

mnist模型部署成功！

---