tensorflow-kubernetes

Utilities and demos of tensorflow machine learning running on Kubernetes.

nodejs-inception-demo

1. Run inception using tensorflow serving

• Build tensorflow serving docker image following instructions here https://www.tensorflow.org/serving/serving_inception (or use mine @ https://hub.docker.com/r/paulwelch/tensorflow-serving-inception/).
• Deploy on Kubernetes

  kubectl apply -f nodejs-inception-demo/inception_serving_k8s.yaml
  

2. Run demo UI to submit an image to inception service

• Build nodejs-inception-demo docker image (or use mine @ https://hub.docker.com/r/paulwelch/image-classifier-ui/).

  cd nodejs-inception-demo
  docker build -t myimages/image-classifier-ui .
  cd ..
  

• Deploy on Kubernetes.

  kubectl apply -f nodejs-inception-demo/inception_ui_k8s.yaml
  

• Depending on your Kubernetes environment, optionally expose the UI service or create an ingress for it. Note: the code is written to run on a /inception path prefix to allow use with path based routing in loadbalancers. Path tested with Traefik.

  for example:
  kubectl apply -f nodejs-inception-demo/inception_ui_k8s_ingress.yaml
  

Authors

Created and maintained by Paul Welch

License

MIT Licensed. See LICENSE for full details.