Skip to content

[Inference Operator] Update CRDs and Helm chart to v3.1#412

Open
zicanl-amazon wants to merge 1 commit intoaws:mainfrom
zicanl-amazon:feature/update-inference-operator-crd-v3.1
Open

[Inference Operator] Update CRDs and Helm chart to v3.1#412
zicanl-amazon wants to merge 1 commit intoaws:mainfrom
zicanl-amazon:feature/update-inference-operator-crd-v3.1

Conversation

@zicanl-amazon
Copy link
Copy Markdown

@zicanl-amazon zicanl-amazon commented Apr 21, 2026

Summary

Update inference operator CRDs and Helm chart to v3.1 (chart version 2.1.0).

Synced from internal AWSCrescendoInferenceOperator repo.

Changes

InferenceEndpointConfig & JumpStartModel CRDs

  • Data capture: 3-tier capture configuration (SageMaker endpoint, LoadBalancer access logs, Model Pod level)
  • DNS automation: Route53 dnsConfig (spec) and dnsStatus (status) for automated DNS record management
  • HuggingFace model source: New huggingFaceModel model source type with modelId, commitSHA, and tokenSecretRef
  • Kubernetes volume source: New kubernetesVolume model source type
  • ServiceAccount support: serviceAccountName field for IRSA credentials on inference pods
  • APS workspace URL pattern fix: Corrected regex pattern for AMP workspace URLs

SageMakerEndpointRegistration CRD

  • Data capture config: sageMakerEndpointDataCaptureConfig for standalone endpoint registrations
  • User-defined tags: tags field in both spec and status for propagating tags to SageMaker resources

Chart.yaml

  • Minor trailing newline fix

values.yaml

  • Added initContainer image configuration (repository and tag)

Files Changed

  • helm_chart/HyperPodHelmChart/charts/inference-operator/config/crd/*.yaml — CRD schema updates
  • helm_chart/HyperPodHelmChart/charts/inference-operator/Chart.yaml — newline fix
  • helm_chart/HyperPodHelmChart/charts/inference-operator/values.yaml — init container config

Validation

  • CRD YAML is valid (verified with YAML parser)
  • Chart.yaml and values.yaml are valid YAML
  • Version numbers are consistent (appVersion=3.1, chart version=2.1.0)
  • Unit tests pass (102/102)
  • helm template — skipped due to pre-existing missing chart dependencies (not related to this change)

Reference

@zicanl-amazon zicanl-amazon requested a review from a team as a code owner April 21, 2026 23:03
@zicanl-amazon zicanl-amazon force-pushed the feature/update-inference-operator-crd-v3.1 branch from 3d3e07d to 08421f9 Compare April 21, 2026 23:25
Sync CRDs, Chart.yaml, and values.yaml from internal inference operator repo.

New features:
- Data capture: 3-tier capture (SageMaker endpoint, LoadBalancer, Model Pod)
- DNS automation: Route53 dnsConfig and dnsStatus
- HuggingFace model source with token secret reference
- Kubernetes volume model source type
- ServiceAccount support for inference pods (IRSA)
- SageMaker endpoint registration: data capture config and user-defined tags
- APS workspace URL pattern fix
- Init container image configuration in values.yaml

Note: pdSpec (disaggregated prefill-decode) excluded, not yet released.
Signed-off-by: Zican Li <zicanl@amazon.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants