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.
90 lines
2.4 KiB
90 lines
2.4 KiB
# Default values for nim.
|
|
# This is a YAML-formatted file.
|
|
# Declare variables to be passed into your templates.
|
|
|
|
# NVIDIA NIMService configuration
|
|
nimService:
|
|
# Name of the NIMService
|
|
name: meta-llama3-8b-instruct
|
|
|
|
# Labels for the NIMService
|
|
labels:
|
|
app.nvidia.com/nim-type: inference
|
|
|
|
# Authentication secret for NVIDIA NGC
|
|
# If existingAuthSecret is set, the secret with this name will be used
|
|
# If empty, a secret will be created with the name specified in authSecret
|
|
existingAuthSecret: ""
|
|
authSecret: ngc-api
|
|
|
|
# Environment variables for the NIMService
|
|
env:
|
|
- name: NIM_GUIDED_DECODING_BACKEND
|
|
value: fast_outlines
|
|
- name: NIM_PEFT_SOURCE
|
|
value: http://nemo-entity-store:8000
|
|
- name: NIM_SERVED_MODEL_NAME
|
|
value: meta/llama-3.2-1b-instruct
|
|
- name: NIM_MODEL_NAME
|
|
value: meta/llama-3.2-1b-instruct
|
|
- name: NIM_PEFT_REFRESH_INTERVAL
|
|
value: "30"
|
|
|
|
# If existingPVC is set, the PVC with this name will be used
|
|
# If empty, a new PVC will be created with the configuration in storage.pvc
|
|
existingPVC: ""
|
|
|
|
# NGC API Key for authentication
|
|
ngcAPIKey: ""
|
|
|
|
# Runtime class name
|
|
runtimeClassName: nvidia
|
|
|
|
# Number of replicas
|
|
replicas: 1
|
|
|
|
# Storage configuration
|
|
storage:
|
|
# PVC configuration for creating a new PVC
|
|
# These settings are used when existingPVC is empty
|
|
pvc:
|
|
storageClass: longhorn
|
|
size: 30Gi
|
|
volumeAccessMode: ReadWriteMany
|
|
# Set to true when using an existing PVC that should be mounted as read-only
|
|
# This is recommended when existingPVC is set
|
|
readOnly: true
|
|
|
|
# Resource requirements
|
|
resources:
|
|
limits:
|
|
nvidia.com/gpu: 1
|
|
|
|
# Service configuration
|
|
expose:
|
|
service:
|
|
type: ClusterIP
|
|
port: 8000
|
|
|
|
# This sets the container image
|
|
image:
|
|
repository: nvcr.io/nim/meta/llama-3.2-3b-instruct
|
|
pullPolicy: IfNotPresent
|
|
tag: "1.8.3"
|
|
|
|
# This is for the secrets for pulling an image from a private repository
|
|
# If existingImagePullSecret is set, the secret with this name will be used
|
|
# If empty, secrets will be created based on the imagePullSecrets configuration
|
|
existingImagePullSecret: ""
|
|
|
|
# Image pull secrets configuration
|
|
# Used to create secrets if existingImagePullSecret is false
|
|
imagePullSecrets:
|
|
- name: nvcrimagepullsecret
|
|
registry: nvcr.io
|
|
username: "$oauthtoken"
|
|
password: ""
|
|
|
|
# This is to override the chart name
|
|
nameOverride: ""
|
|
fullnameOverride: ""
|
|
|