It began as just a simpler way to run TensorFlow jobs on Kubernetes
The Kubeflow project is dedicated to making deployments of machine learning (ML) workflows on Kubernetes simple, portable and scalable. Our goal is not to recreate other services, but to provide a straightforward way to deploy best-of-breed open-source systems for ML to diverse infrastructures.
installation notes on pi cluster
- use 2.1.0 instead of 2.2.0
- kubeflow doesnt work on raspberrypi cluster =(
- clogging my poor longhorn cluster with 20gb pvc claims
- exit kubeflow, switching to alternative MLFlow
investigate kubeflow resources
installs a staggering list of resources into cluster:
- serviceaccounts for kubeflow and ml pipelines
- rbac authorization for everything
- rolebindings
- configmap,services and deployments for
- minio
- cache-server
- metadata-envoy
- metadata-grpc
- minio
- pipeline
- pipelineui
- pipelivisualization
- workflowcontrollermetrics
key points
- uses kserve for model hosting / inference