I am running a Kubernetes cluster on my local machine. In addition, I've managed to set up a Jupyter Hub environment with Kubernetes.
What I want to do is:
- Install Kubeflow Kale on the Jupyter Notebook
- Avoid the full installation of Kubeflow (I don't need everything from there)
- Install the Kubeflow Pipelines as standalone (in the same cluster)
- Deploy my ML models to KFP using Kale (directly from the notebook)
Is that possible? Most online guides (as also mentioned in this post) are requiring the full Kubeflow deployment.
--
What I have already done/tried:
- Followed the official Jupyter Lab guide to set up my own Jupyter Hub
- Followed the official Kubeflow Kale installation guide
- Found some instructions to install Kubeflow Pipelines as standalone