Modelplane Modelplane docs
Version
main v0.3 Latest v0.2 v0.1

InferenceClass Custom Resource

On this page
This document is for an older version of Modelplane.

This document applies to Modelplane v0.2 and not to the latest release v0.3.

Hardware recipe defining GPU type, count, and provisioning for a node pool.

Concept guide: Define Hardware Classes →

#Metadata

API version
modelplane.ai/v1alpha1
Kind
InferenceClass
Scope
Cluster
Short names
icl

#Example

Manifest
apiVersion: modelplane.ai/v1alpha1
kind: InferenceClass
metadata:
  name: gke-l4-1x-g2
spec:
  description: "GKE g2-standard-8, 1x NVIDIA L4"
  provisioning:
    provider: GKE
    gke:
      machineType: g2-standard-8
      diskSizeGb: 100
      accelerator:
        type: nvidia-l4
        count: 1
  devices:
    - name: gpu
      claim: DRA
      driver: gpu.nvidia.com
      deviceClassName: gpu.nvidia.com
      count: 1
      attributes:
        architecture: { string: Ada Lovelace }
      capacity:
        memory: { value: "23034Mi" }

#Spec

# description optional string

Human-readable description of the class.

# devices required object[] 1–16 items
# attributes optional map[string]object

DRA-style typed attributes for this device. Keys are bare names (e.g. architecture); the domain comes from the device’s driver. Each value sets exactly one typed field.

# capacity optional map[string]object

DRA-style capacity quantities for this device. Keys are bare names (e.g. memory); values are Kubernetes Quantities.

# claim optional enum: DRA | Synthetic default: DRA

How Modelplane treats this device. DRA emits it as a request in a ResourceClaim, so DRA binds a matching device to the pod at admission time; use it for hardware a real DRA driver exposes. Synthetic describes the device for fleet scheduling only and never claims it; use it for hardware that matters for placement but has no DRA driver yet, like an InfiniBand fabric.

# count optional integer 1–64 default: 1

How many of this device a node has.

# deviceClassName optional string 1–253 chars

Name of the cluster-scoped DRA DeviceClass to claim this device through. Required for claim: DRA devices; the DRA driver install creates the DeviceClass (e.g. gpu.nvidia.com). Ignored for Synthetic devices.

# driver required string 1–253 chars

DRA driver that owns this device (e.g. gpu.nvidia.com). Becomes the attribute/capacity domain a nodeSelector reads as device.attributes[""]..

# name required string 1–63 chars

Name of this device within the class (e.g. gpu, nic).

# provisioning optional object

How to provision a node pool of this class. Omit for classes that describe BYO node pools that already exist.

# aks optional object
# accelerator required object

GPU accelerator to attach when provisioning the node pool. Provisioning input only: the scheduler matches against spec.devices, not this block.

# count required integer 1–16
# type required string 1–63 chars

GPU accelerator type (e.g. nvidia-a100, nvidia-h100). Informational - reported on the consuming InferenceCluster’s status.

# diskSizeGb optional integer ≥ 10 default: 100
# vmSize required string ≥ 1 chars

Azure VM size (e.g. Standard_NC24ads_A100_v4, Standard_ND96isr_H100_v5). The VM size determines the GPU model and count; the accelerator block below is informational.

# eks optional object
# accelerator required object

GPU accelerator to attach when provisioning the node group. Provisioning input only: the scheduler matches against spec.devices, not this block.

# count required integer 1–16
# type required string 1–63 chars

GPU accelerator type (e.g. nvidia-a10g, nvidia-h100). Informational - reported on the consuming InferenceCluster’s status.

# diskSizeGb optional integer ≥ 10 default: 100
# instanceType required string ≥ 1 chars

EC2 instance type (e.g. g6.xlarge, p4d.24xlarge). The instance family determines the GPU model; the accelerator block below is informational.

# gke optional object
# accelerator required object

GPU accelerator to attach when provisioning the node pool. Provisioning input only: the scheduler matches against spec.devices, not this block, so count here is the GCP machine’s GPU count and need not be restated in devices.

# count required integer 1–16
# type required string 1–63 chars

GPU accelerator type passed to GCP (e.g. nvidia-l4, nvidia-h100-80gb).

# diskSizeGb optional integer ≥ 10 default: 100
# machineType required string ≥ 1 chars
# nebius optional object
# accelerator required object

GPU accelerator to attach when provisioning the node group. Provisioning input only: the scheduler matches against spec.devices, not this block.

# count required integer 1–16
# type required string 1–63 chars

GPU accelerator type (e.g. nvidia-h100, nvidia-l40s). Informational - reported on the consuming InferenceCluster’s status.

# diskSizeGb optional integer ≥ 10 default: 100
# driversPreset optional string 1–63 chars default: cuda13.0

NVIDIA driver stack mk8s preinstalls on the pool’s nodes (e.g. cuda12.4, cuda13.0). Valid values depend on the platform and Kubernetes version, and mk8s validates them, so new presets work without a Modelplane update. Defaults to the only preset mk8s implements on the default Kubernetes version; older presets remain for older versions.

# platform required string 1–63 chars

Nebius compute platform (e.g. gpu-h100-sxm, gpu-l40s-a). Together with preset this determines the GPU model and count; the accelerator block below is informational.

# preset required string 1–63 chars

Resource preset within the platform (e.g. 8gpu-128vcpu-1600gb). Determines the GPU, vCPU, and memory shape of each node.

# provider required enum: GKE | EKS | AKS | Nebius

#Status

# conditions optional object[]