Helm Chart Parameters for NVIDIA NIM for Object Detection

Use this documentation to learn about the helm chart parameters for NVIDIA NIM for Object Detection.

The values that appear are the default values. Refer to the Kubernetes API reference for details on customizing values in the Helm chart.

Deployment Parameters

Name	Description	Value
affinity	Configure pod affinity and anti-affinity.	{}
containerSecurityContext	Configure security settings for the main container.	{}
customCommand	Override the command for the NIM container.	[]
customArgs	Override the arguments for the NIM container.	[]
envVars	Add environment variables to the main container as key-value pairs.	{}
extraVolumes	Add volumes to the deployment.	{}
extraVolumeMounts	Mount volumes from extraVolumes in the main container.	{}
image.repository	Specify the container image repository.	nvcr.io/nim/nvidia/nemotron-object-detection
image.tag	Specify the container image tag.	2.0
image.pullPolicy	Specify the image pull policy.	IfNotPresent
imagePullSecrets	Specify image pull secrets for the main and init containers.	[{name: ngc-secret}]
initContainers	Add init containers to the pod.	{}
nodeSelector	Select nodes on which the pod can run.	{}
podAnnotations	Add annotations to the pod.	{}
podSecurityContext.runAsUser	Specify the user ID for the pod.	1000
podSecurityContext.runAsGroup	Specify the group ID for the pod.	1000
podSecurityContext.fsGroup	Specify the file system group ID for the pod.	1000
replicaCount	Specify the number of deployment replicas.	1
resources.limits.nvidia.com/gpu	Specify the number of GPUs presented to the container.	1
schedulerName	Specify the Kubernetes scheduler.	""
serviceAccount.create	Create a service account.	false
serviceAccount.annotations	Add annotations to the service account.	{}
serviceAccount.name	Specify the service account name.	""
statefulSet.enabled	Deploy the NIM as a StatefulSet.	false
tolerations	Configure pod tolerations.	[{key: nvidia.com/gpu, operator: Exists, effect: NoSchedule}]

Autoscaling Parameters

Values used for autoscaling. If autoscaling is not enabled, these are ignored. They should be overridden on a per-model basis based on quality-of-service metrics as well as cost metrics. This isn’t recommended except with usage of the custom metrics API using something like the prometheus-adapter. Standard metrics of CPU and memory are of limited use in scaling NIM.

Name	Description	Value
autoscaling.enabled	Enable horizontal pod autoscaler.	false
autoscaling.minReplicas	Specify minimum replicas for autoscaling.	1
autoscaling.maxReplicas	Specify maximum replicas for autoscaling.	10
autoscaling.metrics	Array of metrics for autoscaling.	[]

Ingress Parameters

Name	Description	Value
ingress.enabled	Enable ingress.	false
ingress.className	Specify the ingress class name.	""
ingress.annotations	Add ingress annotations.	{}
ingress.hosts[0].host	Specify the ingress host.	chart-example.local
ingress.hosts[0].paths[0].path	Specify the ingress path.	/
ingress.hosts[0].paths[0].pathType	Specify the ingress path type.	ImplementationSpecific
ingress.tls	Specify TLS secrets and hosts.	[]

Probe Parameters

Name	Description	Value
livenessProbe.enabled	Enable the liveness probe.	true
livenessProbe.method	Specify the probe method.	httpGet
livenessProbe.path	Specify the liveness endpoint.	/v1/health/live
livenessProbe.initialDelaySeconds	Specify the initial delay.	15
livenessProbe.timeoutSeconds	Specify the timeout.	1
livenessProbe.periodSeconds	Specify the probe interval.	10
livenessProbe.successThreshold	Specify the success threshold.	1
livenessProbe.failureThreshold	Specify the failure threshold.	3
readinessProbe.enabled	Enable the readiness probe.	true
readinessProbe.method	Specify the probe method.	httpGet
readinessProbe.path	Specify the readiness endpoint.	/v1/health/ready
readinessProbe.initialDelaySeconds	Specify the initial delay.	15
readinessProbe.timeoutSeconds	Specify the timeout.	1
readinessProbe.periodSeconds	Specify the probe interval.	10
readinessProbe.successThreshold	Specify the success threshold.	1
readinessProbe.failureThreshold	Specify the failure threshold.	3
startupProbe.enabled	Enable the startup probe.	true
startupProbe.method	Specify the probe method.	httpGet
startupProbe.path	Specify the startup endpoint.	/v1/health/ready
startupProbe.initialDelaySeconds	Specify the initial delay.	40
startupProbe.timeoutSeconds	Specify the timeout.	1
startupProbe.periodSeconds	Specify the probe interval.	10
startupProbe.successThreshold	Specify the success threshold.	1
startupProbe.failureThreshold	Specify the failure threshold.	180

Storage Parameters

Name	Description	Value
persistence.enabled	Enable persistent volumes.	false
persistence.existingClaim	Specify an existing persistent volume claim.	""
persistence.storageClass	Specify the persistent volume storage class.	""
persistence.accessMode	Specify the persistent volume access mode.	ReadWriteOnce
persistence.stsPersistentVolumeClaimRetentionPolicy.whenDeleted	Specify the retention policy when the StatefulSet is deleted.	Retain
persistence.stsPersistentVolumeClaimRetentionPolicy.whenScaled	Specify the retention policy when the StatefulSet is scaled.	Retain
persistence.size	Specify the persistent volume claim size.	50Gi
persistence.annotations	Add annotations to the persistent volume claim.	{}
hostPath.enabled	Enable a host-path model cache volume.	false
hostPath.path	Specify the model cache path on the node.	/model-store
nfs.enabled	Enable a direct NFS mount.	false
nfs.server	Specify the NFS server.	nfs-server.example.com
nfs.path	Specify the path exported by the NFS server.	/exports
nfs.readOnly	Mount the NFS volume as read-only.	false

Service Parameters

Name	Description	Value
service.name	Override the default service name.	""
service.type	Specify the service type.	ClusterIP
service.httpPort	Specify the HTTP port.	8000
service.serverPort	Specify the server port.	0
service.grpcPort	Specify the gRPC port.	0
service.metricsPort	Specify the metrics port.	0
service.annotations	Add annotations to the service.	{}
service.labels	Add labels to the service.	{}
service.nodePort	Specify the node port for a NodePort service.	0

Related Topics

• Deploy NVIDIA NIM for Object Detection on Kubernetes