$ pip install -r requirements.txt
$ npm install
$ gunicorn main:app --log-file=-
docker build -t submod/mnist-app -f Dockerfile.nodejs .
Ensure that you are connected to an OpenShift project
oc new-project test
Create the template for mnist web application.
oc create -f https://raw.githubusercontent.com/sub-mod/mnist-app/master/template.json
Once the tensorflow serving endpoints are created deploy the mnist-app. This app needs 2 serving endpoints.For more details look here tensorflow-serving-s2i
oc new-app --template=mnistapp --param=APPLICATION_NAME=mnist-app \
--param=PREDICTION_SERVICE1=tf-reg \
--param=PREDICTION_SERVICE2=tf-cnn
The UI layer code was used with permission from Yoshihiro Sugi (@sugyan). The backend was redeveloped to work with tensorflow model server and openshift.