jacobpolloreno/tensorflow_serving_models

Tensorflow Serving files to build, train, test, and export models.

Tensorflow + TF Serving Examples

Setup

Docker

1. Build docker image:

docker build --pull -t $USER/tensorflow-serving-devel -f Dockerfile.devel . 

2. Run the docker container:

docker run --name=tf_container -it $USER/tensorflow-serving-devel

Copy the models into the container

cd <path to project>
docker cp ./export tf_container:/serving

4. Start it again with

docker start -i tf_container

5. Start serving (after you enter docker container)

tensorflow_model_server --port=9000 --model_name=MODELNAME --model_base_path=EXPORTPATH &> logs &

gRPC Client

1. Get container IP address

docker network inspect bridge | grep IPv4Address

2. Test your client

cd <path to project>
python MODEL_client.py --server=<IP ADDRESS>:9000 ...