Add helm inference

helm_chart/HyperPodHelmChart/charts/inference-operator/Chart.yaml

@@ -15,13 +15,11 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 1.0.0
+version: 1.1.0
 
-# This is the version number of the application being deployed. This version number should be
-# incremented each time you make changes to the application. Versions are not expected to
-# follow Semantic Versioning. They should reflect the version the application is using.
-# It is recommended to use it with quotes.
-appVersion: "2.0"
+# This is the version number of the application being deployed. Keep this aligned 
+# with operator image MAJOR.MINOR version.
+appVersion: "2.1"
 
 dependencies:
 - name: aws-mountpoint-s3-csi-driver

helm_chart/HyperPodHelmChart/charts/inference-operator/config/crd/inference.sagemaker.aws.amazon.com_inferenceendpointconfigs.yaml

@@ -696,7 +696,7 @@ spec:
                       l2CacheBackend:
                         description: L2 cache backend type. Required when L2CacheSpec
                           is provided.
-                        pattern: (?i)redis
+                        pattern: (?i)redis|tieredstorage
                         type: string
                       l2CacheLocalUrl:
                         description: Provide the L2 cache URL to local storage
@@ -721,6 +721,12 @@ spec:
                     - round_robin
                     type: string
                 type: object
+              maxDeployTimeInSeconds:
+                default: 3600
+                description: Maximum allowed time in seconds for the deployment to
+                  complete before timing out. Defaults to 1 hour (3600 seconds)
+                format: int32
+                type: integer
               metrics:
                 description: Configuration for metrics collection and exposure
                 properties:
@@ -1617,12 +1623,6 @@ spec:
                     - round_robin
                     type: string
                 type: object
-              maxDeployTimeInSeconds:
-                default: 3600
-                description: Maximum allowed time in seconds for the deployment to
-                  complete before timing out. Defaults to 1 hour (3600 seconds)
-                format: int32
-                type: integer
               metrics:
                 description: Configuration for metrics collection and exposure
                 properties:

helm_chart/HyperPodHelmChart/charts/inference-operator/config/manager/manager.yaml

@@ -48,6 +48,94 @@ spec:
         # versions < 1.19 or on vendors versions which do NOT support this field by default (i.e. Openshift < 4.11 ).
         # seccompProfile:
         #   type: RuntimeDefault
+      initContainers:
+        - command:
+            - bash
+            - -lc
+            - |
+              set -euo pipefail
+              KUBECTL="$(command -v kubectl || true)"
+              if [ -z "${KUBECTL}" ]; then
+                for p in /opt/bitnami/kubectl/bin/kubectl /usr/local/bin/kubectl /usr/bin/kubectl /bin/kubectl; do
+                if [ -x "$p" ]; then KUBECTL="$p"; break; fi
+              done
+              fi
+              if [ -z "${KUBECTL}" ]; then
+                echo "kubectl not found in PATH or common locations" > /dev/termination-log
+                exit 2
+              fi
+
+              CHECKS="${CHECKS:-drivers crds}"
+
+              log() { echo "$1" > /dev/termination-log; }
+
+              require_csidriver() {
+                local provisioner="$1"
+                local friendly="$2"
+
+                # Try with error capture so we can disambiguate RBAC vs missing
+                if "${KUBECTL}" get csidriver "$provisioner" >/dev/null 2>&1 || \
+                   "${KUBECTL}" get csidrivers.storage.k8s.io "$provisioner" >/dev/null 2>&1; then
+                  return 0
+                fi
+
+
+                # Final attempt to capture the real error
+                err_msg="$("${KUBECTL}" get csidriver "$provisioner" 2>&1 || true)"
+                [ -z "$err_msg" ] && err_msg="$("${KUBECTL}" get csidrivers.storage.k8s.io "$provisioner" 2>&1 || true)"
+
+                if echo "$err_msg" | grep -qiE 'forbidden|permission|unauthorized|cannot.*get'; then
+                  log "$friendly check failed: RBAC insufficient to read CSIDriver $provisioner. "${KUBECTL}" said: ${err_msg}"
+                  exit 2
+                fi
+
+                log "$friendly not installed (missing CSIDriver $provisioner). kubectl said: ${err_msg}"
+                exit 1
+              }
+
+              require_crd() {
+                local crd="$1"
+                # Same idea: attempt and parse error text
+                if "${KUBECTL}" get crd "$crd" >/dev/null 2>&1; then
+                  return 0
+                fi
+                err="$("${KUBECTL}" get crd "$crd" 2>&1 || true)"
+                if echo "$err" | grep -qiE 'forbidden|permission|unauthorized|cannot.*get'; then
+                  log "CRD check failed: RBAC insufficient to read $crd. "${KUBECTL}" said: ${err}"
+                  exit 2
+                fi
+                log "Missing required CRD: $crd. "${KUBECTL}" said: ${err}"
+                exit 1
+              }
+
+              # Dispatch selected checks
+              for c in $CHECKS; do
+                case "$c" in
+                  drivers)
+                    require_csidriver "s3.csi.aws.com" "S3 CSI driver"
+                    require_csidriver "fsx.csi.aws.com" "FSx CSI driver"
+                    ;;
+                  crds)
+                    require_crd "certificaterequests.cert-manager.io" "cert-manager CRD"
+                    require_crd "certificates.cert-manager.io" "cert-manager CRD"
+                    ;;
+                  *)
+                    log "Unknown check: $c"
+                    exit 1
+                    ;;
+                esac
+              done
+
+              log "Checks passed: $CHECKS"
+              exit 0
+          env:
+            - name: CHECKS
+              value: "drivers crds"
+          image: "public.ecr.aws/bitnami/kubectl:1.30"
+          imagePullPolicy: Always
+          name: check-csi-drivers
+          resources: { }
+          terminationMessagePath: /dev/termination-log
       containers:
       - command:
         - /hyperpod-inference-manager
@@ -93,7 +181,7 @@ spec:
         resources:
           limits:
             cpu: 500m
-            memory: 128Mi
+            memory: 256Mi
           requests:
             cpu: 10m
             memory: 64Mi
@@ -125,4 +213,4 @@ spec:
       volumes:
       - name: webhook-certs
         secret:
-          secretName: webhook-server-cert
+          secretName: webhook-server-cert

helm_chart/HyperPodHelmChart/charts/inference-operator/values.yaml

@@ -21,7 +21,7 @@ image:
     ap-southeast-4: 311141544681.dkr.ecr.ap-southeast-4.amazonaws.com
     ap-southeast-3: 158128612970.dkr.ecr.ap-southeast-3.amazonaws.com
     eu-south-2: 025050981094.dkr.ecr.eu-south-2.amazonaws.com
-  tag: v2.0
+  tag: v2.1
   pullPolicy: Always
   repository:
 hyperpodClusterArn:

hyperpod-jumpstart-inference-template/hyperpod_jumpstart_inference_template/registry.py

@@ -10,13 +10,17 @@
 # distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
-from hyperpod_jumpstart_inference_template.v1_0 import model as v1
-from hyperpod_jumpstart_inference_template.v1_0.template import TEMPLATE_CONTENT as v1_template
+from hyperpod_jumpstart_inference_template.v1_0 import model as v1_0
+from hyperpod_jumpstart_inference_template.v1_1 import model as v1_1
+from hyperpod_jumpstart_inference_template.v1_0.template import TEMPLATE_CONTENT as v1_0_template
+from hyperpod_jumpstart_inference_template.v1_1.template import TEMPLATE_CONTENT as v1_1_template
 
 SCHEMA_REGISTRY = {
-    "1.0": v1.FlatHPJumpStartEndpoint,
+    "1.0": v1_0.FlatHPJumpStartEndpoint,
+    "1.1": v1_1.FlatHPJumpStartEndpoint,
 }
 
 TEMPLATE_REGISTRY = {
-    "1.0": v1_template
+    "1.0": v1_0_template,
+    "1.1": v1_1_template,
 }

hyperpod-jumpstart-inference-template/hyperpod_jumpstart_inference_template/v1_1/__init__.py

@@ -0,0 +1,12 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.

hyperpod-jumpstart-inference-template/hyperpod_jumpstart_inference_template/v1_1/model.py

@@ -0,0 +1,136 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+from pydantic import BaseModel, Field, model_validator, ConfigDict
+from typing import Optional
+
+# reuse the nested types
+from sagemaker.hyperpod.inference.config.hp_jumpstart_endpoint_config import (
+    Model,
+    SageMakerEndpoint,
+    Server,
+    TlsConfig,
+    Validations,
+)
+from sagemaker.hyperpod.inference.hp_jumpstart_endpoint import HPJumpStartEndpoint
+from sagemaker.hyperpod.common.config.metadata import Metadata
+
+
+class FlatHPJumpStartEndpoint(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    namespace: Optional[str] = Field(
+        default=None, description="Kubernetes namespace", min_length=1
+    )
+
+    accept_eula: bool = Field(
+        False,
+        alias="accept_eula",
+        description="Whether model terms of use have been accepted",
+    )
+
+    metadata_name: Optional[str] = Field(
+        None,
+        alias="metadata_name",
+        description="Name of the jumpstart endpoint object",
+        max_length=63,
+        pattern=r"^[a-zA-Z0-9](-*[a-zA-Z0-9]){0,62}$",
+    )
+
+    model_id: str = Field(
+        ...,
+        alias="model_id",
+        description="Unique identifier of the model within the hub",
+        min_length=1,
+        max_length=63,
+        pattern=r"^[a-zA-Z0-9](-*[a-zA-Z0-9]){0,62}$",
+    )
+
+    model_version: Optional[str] = Field(
+        None,
+        alias="model_version",
+        description="Semantic version of the model to deploy (e.g. 1.0.0)",
+        min_length=5,
+        max_length=14,
+        pattern=r"^\d{1,4}\.\d{1,4}\.\d{1,4}$",
+    )
+
+    instance_type: str = Field(
+        ...,
+        alias="instance_type",
+        description="EC2 instance type for the inference server",
+        pattern=r"^ml\..*",
+    )
+
+    accelerator_partition_type: Optional[str] = Field(
+        None, 
+        alias="accelerator_partition_type", 
+        description="MIG profile to use for GPU partitioning",
+        pattern=r"^mig-.*$",
+    )
+
+    accelerator_partition_validation: Optional[bool] = Field(
+        True,
+        alias="accelerator_partition_validation",
+        description="Enable MIG validation for GPU partitioning. Default is true."
+    )
+
+    endpoint_name: Optional[str] = Field(
+        None,
+        alias="endpoint_name",
+        description="Name of SageMaker endpoint; empty string means no creation",
+        max_length=63,
+        pattern=r"^[a-zA-Z0-9](-*[a-zA-Z0-9]){0,62}$",
+    )
+    tls_certificate_output_s3_uri: Optional[str] = Field(
+        None,
+        alias="tls_certificate_output_s3_uri",
+        description="S3 URI to write the TLS certificate",
+        pattern=r"^s3://([^/]+)/?(.*)$",
+    )
+
+    @model_validator(mode="after")
+    def validate_name(self):
+        if not self.metadata_name and not self.endpoint_name:
+            raise ValueError("Either metadata_name or endpoint_name must be provided")
+        return self
+
+    def to_domain(self) -> HPJumpStartEndpoint:
+        if self.endpoint_name and not self.metadata_name:
+            self.metadata_name = self.endpoint_name
+
+        metadata = Metadata(name=self.metadata_name, namespace=self.namespace)
+
+        model = Model(
+            accept_eula=self.accept_eula,
+            model_id=self.model_id,
+            model_version=self.model_version,
+        )
+        validations = Validations(
+            accelerator_partition_validation=self.accelerator_partition_validation,
+        )
+        server = Server(
+            instance_type=self.instance_type,
+            accelerator_partition_type=self.accelerator_partition_type,
+            validations=validations,
+        )
+        sage_ep = SageMakerEndpoint(name=self.endpoint_name)
+        tls = TlsConfig(
+            tls_certificate_output_s3_uri=self.tls_certificate_output_s3_uri
+        )
+        return HPJumpStartEndpoint(
+            metadata=metadata,
+            model=model,
+            server=server,
+            sage_maker_endpoint=sage_ep,
+            tls_config=tls,
+        )