(gen-ai): JumpStartSageMakerEndpoint not working

[LeonLuttenberger]

Describe the bug

I'm unable to deploy the SageMaker endpoint using the JumpStartSageMakerEndpoint construct.

Expected Behavior

The endpoint gets created.

Current Behavior

CloudFormation throws the following error when trying to create the endpoint:

Could not find model data at s3://jumpstart-cache-prod-us-east-1/huggingface-asr/huggingface-asr-whisper-base/artifacts/inference-prepack/v1.0.1/. (Service: AmazonSageMaker; Status Code: 400; Error Code: ValidationException; Request ID: 8e830934-e61a-4c1e-96e5-1663211204b0; Proxy: null)

Reproduction Steps

import * as cdk from "aws-cdk-lib";
import { Construct } from "constructs";
import {
  JumpStartModel,
  SageMakerInstanceType,
  JumpStartSageMakerEndpoint,
} from "@cdklabs/generative-ai-cdk-constructs";

export class SagemakerJumpStartFmEndpointStack extends cdk.Stack {
  jumpStartEndpoint: JumpStartSageMakerEndpoint;

  constructor(scope: Construct, id: string, props: SagemakerJumpStartFmEndpointStackProps) {
    super(scope, id, props);

    this.jumpStartEndpoint = new JumpStartSageMakerEndpoint(this, "JumpStartEndpoint", {
      model: JumpStartModel.HUGGINGFACE_ASR_WHISPER_BASE_2_0_0,
      instanceType: SageMakerInstanceType.of("ml.g5.2xlarge"),
    });
}

Possible Solution

No response

Additional Information/Context

No response

CDK CLI Version

2.111.0

Framework Version

0.1.11

Node.js Version

21.4.0

OS

Mac OS

Language

Typescript

Language Version

No response

Region experiencing the issue

us-east-1

Code modification

No

Other information

No response

Service quota

• I have reviewed the service quotas for this construct

[krokoko]

Hi @LeonLuttenberger, thank you for reporting this ! I was able to reproduce

Issue:
The construct is expecting to get the model artifacts as a single gzip compressed tar archive (.tar.gz suffix)
However, for some models, it is not the case. It seems to affect all models in version 2_0_0 or above.

The construct needs to handle this edge case