keras-serving
keras->tensorflow+grpc+docker=>nodejs
- example of bringing a keras model to production using tensorflow serving
- using custom XOR model with tensor.proto dimensions example
- building & training of the model works with python2.7 on the workstation
- exported model is served via grpc in a C++ server using a Docker-Container
- a nodejs server wraps the grpc api for a simple http POST endpoint
- also ships an advanced multi-model face (emotion, gender) detection example
/face-recog - and and advanced google bigquery (as dataaset) example
/bigquery
Overview
- XOR Setup below
- Face-Recog Setup
- BigQuery Setup
Workflow (Unix - testen on Ubuntu 16.04 64bit)
You will need
python
pip
docker (docker-compose)
1. Install Requirements
./prepare.sh
# installs python dependencies via pip
# builds the docker image for tensorflow_serving (takes a while ~ 30 minutes)
# image size ~ 3.5 GB
checkout build troubleshoot if you are having trouble
2. Build, Train and Serialise Keras Model
python train.py
# results will be in (/result)
3. Load and export Model as Tensorflow Graph
python export.py
# results will be in (/export)
4. Build & Run Containers via docker-compose
./start-servers.sh
# ./stop-servers.sh
5. Test API via curl
curl -X POST \
http://localhost:8080/predict-xor \
-H 'cache-control: no-cache' \
-H 'content-type: application/json' \
-d '{"inputs": [0,1]}'