model-version.yaml

apiVersion: ml.k8s-lab.io/v1alpha1
kind: ModelVersion
metadata:
  name: qwen2.5-7b-k8s-sft-lora-v1
spec:
  owner: platform-llm
  family: qwen2.5
  task: chat-assistant
  stage: candidate
  lineage:
    baseModel:
      name: "Qwen/Qwen2.5-7B"
      revision: "main"
    dataset:
      name: "llm-k8s-sft"
      version: "v1.3.2"
    trainingCode:
      repo: "git@github.com:example/k8s-lab.git"
      commit: "4b7e2c1"
    run:
      id: "sft-qwen2.5-7b-k8s-001"
  artifacts:
    checkpoint:
      uri: "s3://ml-assets/runs/sft-qwen2.5-7b-k8s-001/checkpoint-000120"
    adapter:
      uri: "s3://ml-assets/adapters/qwen2.5-7b-k8s-sft-v1"
    mergedModel:
      uri: "s3://ml-assets/merged/qwen2.5-7b-k8s-sft-v1"
    runtimePackage:
      uri: "s3://ml-assets/runtime/qwen2.5-7b-k8s-sft-v1-awq"
  metrics:
    evalLoss: 1.82
    domainAccuracy: 0.87
    hallucinationRate: 0.06
    safetyRefusalPassRate: 0.98
  approvals:
    dataGovernance: approved
    securityScan: passed
    qa: pending