11. Morpheus-ai-engine

A Helm chart for deploying the NVIDIA Morpheus AI Engine which includes the Triton Inference Server.

11.1. Install Nvidia Morpheus

NOTE: Morpheus AI Engine requires you sign up for an NGC api key, you can sign up here for an api key: Nvidia NGC

git clone https://github.com/graphistry/graphistry-helm && cd graphistry-helm
helm upgrade -i morpheus ./charts/morpheus-ai-engine --namespace morpheus --create-namespace

11.2. Configuration

The following table lists the configurable parameters of the Morpheus-ai-engine chart and their default values.

Parameter

Description

Default

replicaCount

1

ngc.username

"$oauthtoken"

ngc.apiKey

""

ngc.org

""

ngc.team

""

aiengine.registry

"nvcr.io/nvidia"

aiengine.image

"tritonserver"

aiengine.version

"22.06-py3"

aiengine.args

["tritonserver", "--model-repository", "/common/triton-model-repo", "--model-control-mode", "explicit"]

aiengine.resources.limits.nvidia.com/gpu

1

broker.registry

"docker.io"

broker.image

"bitnami/kafka"

broker.version

"2.7.0"

broker.brokerPort

30092

zookeeper.registry

"docker.io"

zookeeper.image

"zookeeper"

zookeeper.version

"3.6.3"

hostCommonPath

"/opt/morpheus/common"

imagePullPolicy

"IfNotPresent"

imagePullSecrets

[{"name": "nvidia-registrykey-secret"}]

serviceAccount.create

false

serviceAccount.name

"morpheus"

platform.openshift

false

loadBalancer.enabled

false

nodeSelector

{}

For more information on NVIDIA Morpheus, visit the Morpheus documentation: NVIDIA Morpheus Documentation