使用docker封装部署模型为微服务
#安装docker
yum install docker
#开启docker服务
service docker start
使用docker拉取tensorflow-serving镜像。
docker pull tensorflow/serving
MODEL_PATH="/data/django-uwsgi/text_labeled/model_train"
docker run -t --rm -p 8501:8501 -v “$MODEL_PATH/movie:/models/movie” -e MODEL_NAME=movie tensorflow/serving &
-p 8501:8501 #http tcp
报错:
模型原保存文件路径:
save_model_path="./movie/"
修改为如下保存方式,解决:
time_ = str(int(time.time()))
save_model_path="./movie/" +time_
(tensorf) [root@bhs model_train]# curl -d '{"instances": [[1.0, 2.0, 5.0]]}' -X POST http://localhost:8501/v1/models/movie:predict
输出:
{
"predictions": [[5.0355615e-20]
]