Wuxinshui

基础服务系列-安装TensorFlow Serving

2019/12/31

TensorFlow Serving is a flexible, high-performance serving system for machine learning models, designed for production environments.

下载镜像

docker pull tensorflow/serving

网络原因，可能会导致timeout，多尝试几次。

clone代码

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

启动镜像

docker run -t --rm -p 8501:8501 \
    -v "/root/tf-serving/serving/tensorflow_serving/servables/tensorflow/testdata/saved_model_half_plus_two_cpu:/models/half_plus_two" \
    -e MODEL_NAME=half_plus_two \
    tensorflow/serving &_model_half_plus_two_cpu:/models/half_plus_two" \
    -e MODEL_NAME=half_plus_two \
    tensorflow/serving &

启动后。。。

测试

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

测试成功。。。

遇到的问题

No versions of servable half_plus_two found under base path /models/half_plus_two

解决方案

创建目录：/models/half_plus_two

参考

Serve a Tensorflow model in 60 seconds

作者：Wuxinshui
出处：http://wuxinshui.github.io
版权归作者所有，转载请注明出处

Post Directory