본문으로 건너뛰기

ModelDeployment

Container deployment settings for an internally hosted model.

imageImage (string)필수

Fully qualified container image reference. Custom images must be digest-pinned and cannot use the latest tag.

artifact_version object

Artifact version identifier to deploy. Omit it to resolve the latest version when deployment starts.

하나 이상(anyOf)
string
command object

Optional container entrypoint override.

하나 이상(anyOf)
  • Array [
    • string
  • ]
    • args object

    Optional arguments that override the image command.

    하나 이상(anyOf)
  • Array [
    • string
  • ]
    • portPort (integer)
    기본값: 8080
    cpu object

    Requested CPU amount, such as 500m.

    하나 이상(anyOf)
    string
    memory object

    Requested memory amount, such as 2Gi.

    하나 이상(anyOf)
    string
    gpu object

    GPU count request

    하나 이상(anyOf)
    integer
    replicasReplicas (integer)

    Number of model instances to run. Higher values increase resource usage.

    기본값: 1
    health_pathHealth Path (string)

    HTTP path used to check whether the model is ready and healthy.

    기본값: /healthz
    env object

    Non-secret environment variables passed to the model container. Store secrets separately.

    property name*string
    node_selector object

    Node labels that every selected execution node must match.

    property name*string
    tolerations object[]

    Node taints that the model deployment is allowed to tolerate.

  • Array [
    • key object

    Taint key matched by this toleration, or null to match every key when operator is Exists.

    하나 이상(anyOf)
    string
    operatorOperator (string)

    'Equal' requires value; 'Exists' ignores it.

    Possible values: [Exists, Equal]

    기본값: Equal
    value object
    하나 이상(anyOf)
    string
    effect object

    Taint effect matched by this toleration, or null to match every effect.

    하나 이상(anyOf)
    string

    Possible values: [NoSchedule, PreferNoSchedule, NoExecute]

    toleration_seconds object

    For effect='NoExecute', evict after this many seconds.

    하나 이상(anyOf)
    integer
  • ]
    • ModelDeployment
    {
      "image": "string",
      "artifact_version": "string",
      "command": [
        "string"
      ],
      "args": [
        "string"
      ],
      "port": 8080,
      "cpu": "string",
      "memory": "string",
      "gpu": 0,
      "replicas": 1,
      "health_path": "/healthz",
      "env": {},
      "node_selector": {},
      "tolerations": [
        {
          "key": "string",
          "operator": "Equal",
          "value": "string",
          "effect": "NoSchedule",
          "toleration_seconds": 0
        }
      ]
    }