Documentation: docs.nebuly.com/nos/overview

If you like the project please support it by leaving a star ✨

nos is the open-source module to efficiently run AI workloads on Kubernetes, increasing GPU utilization, cutting down infrastructure costs and improving workloads performance.

Currently, the available features are:

• Dynamic GPU partitioning: allow to schedule Pods requesting fractions of GPU. GPU partitioning is performed automatically in real-time based on the Pods pending and running in the cluster, so that Pods can request only the resources that are strictly necessary and GPUs are always fully utilized.

• Elastic Resource Quota management: increase the number of Pods running on the cluster by allowing namespaces to borrow quotas of reserved resources from other namespaces as long as they are not using them.

• NVIDIA GPU Operator
• Nebuly k8s-device-plugin (optional, required only if you want to enable MPS partitioning)
• cert-manager (optional, but recommended)

You can install nos using Helm 3 (recommended). You can find all the available configuration values in the Chart documentation.

helm install oci://ghcr.io/nebuly-ai/helm-charts/nos \
  --version 0.1.0 \
  --namespace nebuly-nos \
  --generate-name \
  --create-namespace

Alternatively, you can use Kustomize by cloning the repository and running make deploy.

Join the community | Contribute