Messing with distributed TensorFlow and Kubernetes cluster.
- distributed training: chestxray
While writing training yaml files we understood why really kubeflow exists... Configuring workers environnements and dns services is not complicated but time consuming. It can be automatically generated ! We'll continue messing with all this stuff in order to determine weither we should use kubeflow when its stable (version >= 1) or create our own generation tool.
- Installed and configured
gcloud(verify that you have enough quotas for using GKE) - Installed
kubectl
See ./docs