Repository for dip
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

504 lines
18 KiB

# The chart generates top-level secrets (custom-values.yaml overrides these)
# -- You can use an existing Kubernetes secret for pulling images. The chart uses the `ngcAPIKey` value to generate the secret if you leave this empty.
existingSecret: ""
# -- You can specify an existing Kubernetes image pull secret for pulling images from the container registry. The chart doesn't reference any image pull secret if you leave this empty.
existingImagePullSecret: ""
# -- Your NVIDIA GPU Cloud (NGC) API key authenticates and enables pulling images from the NGC container registry. The existing secret overrides this key if you provide one to the `existingSecret` key.
ngcAPIKey: YOUR-NGC-API-KEY
# -- List of image pull secrets. Existing secrets override these values if you specify them. Use this only for experimentation when you want to hardcode a secret in your values file.
# @default -- `[{"name":"nvcrimagepullsecret","password":"YOUR-NGC-API-KEY","registry":"nvcr.io","username":"$$oauthtoken"}]`
imagePullSecrets:
- name: nvcrimagepullsecret
registry: nvcr.io
username: $oauthtoken
password: YOUR-NGC-API-KEY
global:
# -- The name of the image pull secret to use globally across all services.
imagePullSecrets:
- name: nvcrimagepullsecret
data-store:
# -- Specifies whether to install the NeMo Data Store microservice.
enabled: true
# uncomment this section when ingress is enabled
# external:
# rootUrl: https://nemo-datastore.example.com
# domain: nemo-datastore.example.com
# -- The number of replicas for the NeMo Data Store microservice deployment.
replicaCount: 1
persistence:
# -- Specifies whether to enable persistent volume for the NeMo Data Store microservice.
enabled: true
# -- The name of the persistent volume claim.
claimName: datastore-shared-storage
# -- The size of the persistent volume.
size: 1Gi
# -- The storage class to use for the persistent volume.
storageClass: ""
# -- The access modes for the persistent volume.
accessModes:
- ReadWriteOnce
objectStore:
# -- Specifies whether to enable object storage for the NeMo Data Store microservice.
enabled: false
serviceAccount:
# -- Specifies whether to create a service account for the NeMo Data Store microservice.
create: true
# -- The service account name for the NeMo Data Store microservice.
name: gitea
# -- The service name for the NeMo Data Store microservice.
serviceName: nemo-data-store
strategy:
# -- The type of deployment strategy to use for the NeMo Data Store microservice.
type: Recreate
customizer:
# -- Specifies whether to install the NeMo Customizer microservice.
enabled: true
postgresql:
# -- The name override for the PostgreSQL database.
nameOverride: customizerdb
serviceAccount:
name: customizer-postgresql
# -- Specifies whether to create a new service account for PostgreSQL.
create: true
customizerConfig:
# -- Specifies the internal K8s DNS record for the NeMo Data Store service.
nemoDataStoreURL: "http://nemo-data-store:3000"
# -- Specifies the internal K8s DNS record for the NeMo Entity Store service.
entityStoreURL: "http://nemo-entity-store:8000"
# -- Specifies the internal K8s DNS record for the MLflow service.
mlflowURL: ""
# -- Specifies the model information to configure with the NeMo Customizer microservice. To find a complete list of models that NeMo Customizer supports, see [Model Catalog](https://docs.nvidia.com/nemo/microservices/latest/fine-tune/models/index.html).
models:
meta/llama-3.1-8b-instruct:
# -- Specifies whether to use the `meta/llama-3.1-8b-instruct` model for a customization job.
enabled: true
training:
# -- Specifies where to store the customization job information during training.
pvc:
storageClass: ""
serviceName: nemo-customizer
# -- Specifies whether to enable the PVC for storing base models.
modelsStorage:
enabled: true
# -- Specifies the storage class to use for the PVC.
storageClassName: ""
# -- Specifies the size of the PVC.
size: 20Gi
# -- Specifies the access modes for the PVC.
accessModes:
- ReadWriteOnce
workspaceStorage:
# -- Specifies whether to enable the PVC for storing a training job's workspace.
enabled: false
# -- Specifies the image pull secret for the NeMo Data Store tools.
nemoDataStoreTools:
imagePullSecret: nvcrimagepullsecret
nim:
# -- Specifies whether to deploy a NIM for LLM during the Helm installation of the chart. You can deploy a single static NIM by enabling this object and its values. When enabled, the chart deploys `meta/llama-3.1-8b-instruct` as the default NIM.
enabled: false
image:
# Adjust to the actual location of the image and version you want
repository: nvcr.io/nim/meta/llama-3.1-8b-instruct
tag: "1.8.3"
imagePullSecrets:
- name: nvcrimagepullsecret
model:
# -- The name of the model to deploy as NIM.
name: meta/llama-3.1-8b-instruct # not strictly necessary, but enables running "helm test"
# -- The NGC API secret for model access.
ngcAPISecret: ngc-api
service:
# -- Labels for the NIM service.
labels:
app.nvidia.com/nim-type: inference
# -- Environment variables for the NIM service.
env:
- name: NIM_PEFT_SOURCE
value: http://nemo-entity-store:8000
- name: NIM_PEFT_REFRESH_INTERVAL
value: "30"
- name: NIM_MAX_CPU_LORAS
value: "16"
- name: NIM_MAX_GPU_LORAS
value: "8"
persistence:
# -- Specifies whether to enable persistence volume claim (PVC) for the NIM service.
enabled: true
# -- Specifies the storage class to use for the PVC.
storageClass: ""
# -- Annotations for the PVC.
annotations:
helm.sh/resource-policy: keep
statefulSet:
# -- Specifies whether to enable a stateful set for the NIM service.
enabled: false
# -- Specifies resources for the NIM service.
resources:
limits:
nvidia.com/gpu: 1
# -- Specifies requests for the NIM service.
requests:
nvidia.com/gpu: 1
evaluator:
# -- Specifies whether to install the NeMo Evaluator microservice.
enabled: true
argoWorkflows:
# -- Specifies whether to enable the default Argo Workflows integration for the NeMo Evaluator microservice.
enabled: true
# -- The Argo Workflows service name for the NeMo Evaluator microservice.
serviceName: argo-workflows-server
server:
authModes:
- "server"
# -- The service port for the Argo Workflows server.
servicePort: 2746
# -- Specifies whether to secure the connection to the Argo Workflows server.
secure: true
serviceAccount:
# -- The name of the service account for the Argo Workflows server.
name: "argo-workflows-server"
crds:
# -- Specifies whether to install the CRDs for Argo Workflows.
install: false
controller:
serviceAccount:
# -- The name of the service account for the Argo Workflows controller.
name: "argo-workflows-controller"
external:
# -- The external URL of the NeMo Data Store microservice.
dataStore:
endpoint: http://nemo-data-store:3000/v1/hf
milvus:
standalone:
persistence:
persistentVolumeClaim:
# -- Specifies the storage class to use for the Milvus PVC.
storageClass: ""
postgresql:
# -- The name override for the default PostgreSQL database.
nameOverride: evaluatordb
serviceAccount:
# -- The name of the service account for PostgreSQL.
name: evaluator-postgresql
primary:
networkPolicy:
# -- Specifies whether to enable the network policy for the PostgreSQL service.
enabled: false
serviceName: nemo-evaluator
guardrails:
# -- Specifies whether to install the NeMo Guardrails microservice.
enabled: true
# -- The service name for the NeMo Guardrails microservice.
serviceName: nemo-guardrails
# -- The environment variables for the NeMo Guardrails microservice.
env:
# -- The NIM endpoint URL for the NeMo Guardrails microservice.
NIM_ENDPOINT_URL: http://nemo-nim-proxy:8000/v1
nemo-operator:
# -- Specifies whether to install the NeMo Operator microservice.
enabled: true
controllerManager:
manager:
# -- The scheduler to use for the NeMo Operator. Available options: `volcano` and `runai`.
scheduler: "volcano"
nim-operator:
# -- Specifies whether to install the NIM Operator.
enabled: true
nfd:
nodeFeatureRules:
# -- Specifies whether to enable device ID feature rules.
deviceID: false
dgxc-admission-controller:
# -- Specifies whether to install the DGX Cloud Admission Controller.
enabled: false
kubernetesAffinity:
# -- Specifies whether to enable Kubernetes affinity.
enabled: false
enableDGXCRbac: false
# -- Specifies whether to enable network policies.
enableNetworkPolicies: false
# -- Specifies whether to enable integrations support.
enableIntegrationsSupport: false
# -- Specifies whether to enable the labeler.
enableLabeler: false
# -- The image pull secret for the DGX Cloud Admission Controller.
imagePullSecret: nvcrimagepullsecret
dgxcController:
# -- The image for the DGX Cloud Admission Controller.
image: "nvcr.io/nvidia/nemo-microservices/dgxc-admission-controller:1.642.0"
cloudProviderSpec:
# -- The type of the dummy cloud provider.
dummy:
type: dummy
entity-store:
# -- Specifies whether to install the NeMo Entity Store microservice.
enabled: true
postgresql:
# -- The name override for the PostgreSQL database.
nameOverride: entity-storedb
serviceAccount:
# -- The service account name for PostgreSQL.
name: entity-store-postgresql
# -- Specifies whether to create a new service account for PostgreSQL.
create: true
appConfig:
# -- The base URL for the NIM service used by the NeMo Entity Store microservice.
BASE_URL_NIM: http://nemo-nim-proxy:8000
# -- The base URL for the NeMo Data Store microservice.
BASE_URL_DATASTORE: http://nemo-data-store:3000/v1/hf
# -- The service name for the NeMo Entity Store microservice.
serviceName: nemo-entity-store
volcano:
# -- Specifies whether to enable the default Volcano scheduler installation. To learn more, see [Volcano](https://docs.nvidia.com/nemo/microservices/latest/set-up/deploy-as-microservices/customizer.html#volcano).
enabled: false
deployment-management:
# -- Specifies whether to install the NeMo Deployment Management microservice.
enabled: true
image:
# -- The pull policy for the NeMo Deployment Management container image.
pullPolicy: IfNotPresent
deployments:
# -- The URL for the NIM PEFT source.
nimPeftSource: http://nemo-entity-store:8000
# -- The URL for the NeMo Entity Store service.
entityStoreUrl: http://nemo-entity-store:8000
# -- The period in seconds for model synchronization.
modelSyncPeriod: "30"
# -- The image pull secrets for the NeMo Deployment Management microservice.
nimImagePullSecrets: ["nvcrimagepullsecret"]
# -- The service name for the NeMo Deployment Management microservice.
serviceName: nemo-deployment-management
nim-proxy:
# -- Specifies whether to install the NIM Proxy microservice.
enabled: true
monitoring:
# -- Specifies whether to enable monitoring for the NIM Proxy microservice.
enabled: false
# -- The service name for the NIM Proxy microservice.
serviceName: nemo-nim-proxy
ingress:
# -- Specifies whether to enable the ingress.
enabled: false
# -- Annotations for the ingress resource.
annotations: {}
# -- The ingress class to use if your cluster has more than one class.
className: ""
# -- TLS configurations.
tls: []
# -- A map of hosts and their corresponding paths for the ingress.
hosts:
default:
# -- The host name for the default ingress host.
name: ""
# -- The path rules for the default ingress host.
paths:
- path: /v1/namespaces
pathType: Prefix
service: nemo-entity-store
port: 8000
- path: /v1/projects
pathType: Prefix
service: nemo-entity-store
port: 8000
- path: /v1/datasets
pathType: Prefix
service: nemo-entity-store
port: 8000
- path: /v1/repos
pathType: Prefix
service: nemo-entity-store
port: 8000
- path: /v1/models
pathType: Prefix
service: nemo-entity-store
port: 8000
- path: /v1/customization
pathType: Prefix
service: nemo-customizer
port: 8000
- path: /v1/evaluation
pathType: Prefix
service: nemo-evaluator
port: 7331
- path: /v1/guardrail
pathType: Prefix
service: nemo-guardrails
port: 7331
- path: /v1/deployment
pathType: Prefix
service: nemo-deployment-management
port: 8000
nimProxy:
# -- The host name for the second ingress host for the NIM Proxy microservice.
name: nim.test
# -- The path rules for the second ingress host.
paths:
- path: /v1/completions
pathType: Prefix
service: nemo-nim-proxy
port: 8000
- path: /v1/chat
pathType: Prefix
service: nemo-nim-proxy
port: 8000
- path: /v1/models
pathType: Prefix
service: nemo-nim-proxy
port: 8000
dataStore:
# -- The host name for the third ingress host for the NeMo Data Store microservice.
name: data-store.test
# -- The path rules for the third ingress host.
paths:
- path: /
pathType: Prefix
service: nemo-data-store
port: 3000
# -- Specifies whether to enable the virtual service. If you are not using istio and virtualservices, it can be useful to create some virtual services for the NeMo Microservices system. Don't enable unless you use istio.
# @default -- A virtual service configuration template.
virtualService:
# -- Specifies whether to enable the virtual service.
enabled: false
# -- Labels for the virtual service.
labels: {}
# -- Annotations for the virtual service.
annotations: {}
main:
# -- A list of gateways for the virtual service.
gateways: []
# -- A list of hosts for the virtual service.
hosts: []
entries:
customizer:
# -- The CORS policy for the virtual NeMo Customizer service.
corsPolicy: {}
# -- The match for the virtual NeMo Customizer service.
match:
- uri:
prefix: "/v1/customization"
# -- The route for the virtual NeMo Customizer service.
route:
- destination:
host: nemo-customizer
port:
number: 8000
deployment-management:
# -- The CORS policy for the virtual NeMo Deployment Management service.
corsPolicy: {}
# -- The match for the virtual NeMo Deployment Management service.
match:
- uri:
prefix: "/v1/deployment"
# -- The route for the virtual NeMo Deployment Management service.
route:
- destination:
host: nemo-deployment-management
port:
number: 8000
entity-store:
# -- The CORS policy for the virtual NeMo Entity Store service.
corsPolicy: {}
# -- The match for the virtual NeMo Entity Store service.
match:
- uri:
prefix: "/v1/namespaces"
- uri:
prefix: "/v1/projects"
- uri:
prefix: "/v1/datasets"
- uri:
prefix: "/v1/repos"
- uri:
prefix: "/v1/models"
# -- The route for the virtual NeMo Entity Store service.
route:
- destination:
host: nemo-entity-store
port:
number: 8000
evaluator:
# -- The CORS policy for the virtual NeMo Evaluator service.
corsPolicy: {}
# -- The match for the virtual NeMo Evaluator service.
match:
- uri:
prefix: "/v1/evaluation"
# -- The route for the virtual NeMo Evaluator service.
route:
- destination:
host: nemo-evaluator
port:
number: 7331
guardrails:
# -- The CORS policy for the virtual NeMo Guardrails service.
corsPolicy: {}
# -- The match for the virtual NeMo Guardrails service.
match:
- uri:
prefix: "/v1/guardrail"
# -- The route for the virtual NeMo Guardrails service.
route:
- destination:
host: nemo-guardrails
port:
number: 7331
additional:
# -- Additional virtual service configurations.
data-store:
# -- The gateways for the virtual NeMo Data Store service.
gateways: []
# -- The hosts for the virtual NeMo Data Store service.
hosts: []
# -- The entries for the virtual NeMo Data Store service.
entries:
data-store:
# -- The CORS policy for the virtual NeMo Data Store service.
corsPolicy: {}
# -- The match for the virtual NeMo Data Store service.
match:
- uri:
prefix: /
# -- The route for the virtual NeMo Data Store service.
route:
- destination:
host: nemo-data-store
port:
number: 3000
nim-proxy:
# -- The gateways for the virtual NIM Proxy service.
gateways: []
# -- The hosts for the virtual NIM Proxy service.
hosts: []
entries:
nim-proxy:
# -- The CORS policy for the virtual NIM Proxy service.
corsPolicy: {}
# -- The match for the virtual NIM Proxy service.
match:
- uri:
prefix: /
# -- The route for the virtual NIM Proxy service.
route:
- destination:
host: nemo-nim-proxy
port:
number: 8000