Serverless inference tutorial using sagemaker sdk

[dwarez]

Note

Low Risk
Documentation-only addition (Jupyter notebook); no production code or SDK behavior changes.

Overview
Adds a new SageMaker SDK v3 tutorial notebook under docs/sagemaker/notebooks/sagemaker-sdk/serverless-inference-embedding-tei/ that walks through serverless (scale-to-zero) inference for a low-traffic semantic-search use case.

The notebook explains when serverless fits vs provisioned real-time endpoints, then deploys BAAI/bge-small-en-v1.5 with the TEI CPU container via ModelBuilder and ServerlessInferenceConfig (memory, max concurrency, optional provisioned concurrency) instead of an instance_type. It covers session/role setup, env-driven sizing, invocation with cold vs warm latency timing, a matplotlib bar chart of cold-start overhead, a cosine-similarity retrieval smoke test, and optional cleanup of endpoint, config, and model resources.

^{Reviewed by Cursor Bugbot for commit 059a64e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit b857108. Configure here.}

[cursor]

Missed SageMaker not-found errors

Low Severity

ignore_not_found only treats ResourceNotFound / ResourceNotFoundException or messages containing not exist as missing resources. SageMaker delete calls for endpoints, configs, and models often raise ValidationException with a Could not find … message, which this helper treats as fatal and re-raises.

 

^{Reviewed by Cursor Bugbot for commit b857108. Configure here.}