Installation

Nobody wants complicated installations. We neither. For this reason we provided Terraform scripts in the past to install fully functional Os4ML on a k3d cluster.

k3d cluster

k3d says it's 'a lightweight wrapper to run k3s (Rancher Lab’s minimal Kubernetes distribution) in docker'. Well yes, that's what it does, and it does it great. We use k3d for daily development. Thus, you can assume that Os4ML runs stably on k3d.

With k3d you can do it like this, for example

git clone https://github.com/WOGRA-AG/terraform-kustomization-os4ml.git
cd terraform-kustomization-os4ml
k3d cluster create --config ./k3d-default.yaml
terraform init
terraform apply -auto-approve

This will install ArgoCD in the cluster and then ArgoCD will then take care of installing Os4ml on its own.

GPU Support

For sure, if you want to do machine learning, you want to use GPUs. So, what about GPU support?

Os4ML automatically uses GPUs if they are available in the Kubernetes cluster. Unfortunately, GPUs are known to be a topic of their own. k3d describes it in their documentation here. If you have problems building the Docker image, remember to install the nvidia-container-toolkit in addition to the nvidia-container-runtime.