charts/kokoro-fastapi/examples/gpu-operator-values.yaml

# Follow the official NVIDIA GPU Operator documentation
#   to install the GPU operator with these settings:
#     https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/getting-started.html
#
# This example is for a Nvidia T4 16gb GPU node pool with only 1 GPU on each node on Azure AKS.
# It uses time-slicing to share the a and claim to the system that 1 GPU is 4 GPUs.
# So each pod has access to a smaller gpu with 4gb of memory.
#
devicePlugin: # Remove this if you dont want to use time-slicing
  config:
    create: true
    name: "time-slicing-config"
    default: "any"
    data:
      any: |-
        version: v1
        flags:
          migStrategy: none
        sharing:
          timeSlicing:
            resources:
            - name: nvidia.com/gpu
              replicas: 4

daemonsets:
  tolerations:
    - key: "sku"
      operator: Equal
      value: "gpu"
      effect: NoSchedule
    - key: "kubernetes.azure.com/scalesetpriority"
      operator: Equal
      value: "spot"
      effect: NoSchedule

node-feature-discovery:
  master:
    tolerations:
      - key: "sku"
        operator: Equal
        value: "gpu"
        effect: NoSchedule
      - key: "kubernetes.azure.com/scalesetpriority"
        operator: Equal
        value: "spot"
        effect: NoSchedule
  worker:
    tolerations:
      - key: "sku"
        operator: Equal
        value: "gpu"
        effect: NoSchedule
      - key: "kubernetes.azure.com/scalesetpriority"
        operator: Equal
        value: "spot"
        effect: NoSchedule