Note Kubeflow

From ESS-WIKI
Jump to: navigation, search

Introduce

Kubeflow is a set of pods that deploy machine learning toolkit in Kubernetes. It help you to make ML training/serving easiler in Kubernetes.

Topics

  1. To enable Kubeflow, Install kubeflow with ksonnet
  2. To make a training with web UI editor, Jupyter Notebooks in Kubernetes
  3. To speed up your training with GPU, Add Nvidia device plugin
  4. To speed up your training with cluster, Distributed Tensorflow in Kubernetes
  5. To export your training model for prediction or inference (TensorFlow-Serving), K8S TF Serving with yaml
  6. To export your training model for prediction or inference (Kubeflow/Ksonnet), K8S TF Serving with ksonnet

Reference