This repo hosts a set of multi-arch images to be run as notebook servers on kubeflow. The images are based on the (IBM Powerai images)[https://hub.docker.com/r/ibmcom/powerai]
This repo follows the naming convention layed out by the (IBM Powerai Docker repoistory)[https://hub.docker.com/r/ibmcom/powerai].
Dockerfile-----
powerai-version - The version of PowerAI installed in image latest is 1.6.2 Available options - 1.6.2, 1.6.2ea (early access)
framework - The framework installed on the image, (options vary depending on image version) Available options - all, tensorflow(>=1.6.0), tensorflow-serving (>=1.6.1), pytorch(>=1.6.0), caffe(>=1.6.0), snapml(>=1.6.0, ppc64le only), xgboost(>=1.6.1, ppc64le only, >=1.6.2, x86_64 and ppc64le), rapids (>=1.6.2, ppc64le only)
python - The python version used by frameworks Available options - py36(>=1.6.2), py37 (>=1.6.2)
architecture - Images are built to support x86_64 and ppc64le architectures (options vary depending on image version) Available options - ppc64le, x86_64
examples Dockerfile-1.6.2-all-py37-ppc64le #Docker image that builds all (including rapids) frameworks for python37 on ppc64le
See how to use Kaniko https://github.com/GoogleContainerTools/kaniko
Get a docker config file which can be created by doing a docker login. You will find it at $HOME/.docker/config.json
Take this file and create a secret.
kubectl create secret generic kaniko-secret --from-file=$HOME/.docker/config.json
These instructions also assume an open access github repo.