#containers#sagemaker#inference#deployment#hugging-face

Serving Image Selection

Independent PiSkill directory guide. The original skill remains hosted by Hugging Face Skills.

What is Serving Image Selection?

Helps choose an appropriate serving container for a Hugging Face model deployment on SageMaker and identify the current container image URI for the target setup.

What does Serving Image Selection do?

Serving Image Selection is a Hugging Face skill for choosing the correct SageMaker inference container family for a Hugging Face model. It distinguishes text-generation LLMs, multimodal models, embeddings, rerankers, diffusion models and other tasks so a deployment uses a compatible and current serving image instead of a guessed container.

Who is Serving Image Selection best for?

  • Teams deploying Hugging Face models on SageMaker
  • Developers choosing between vLLM, TEI and other inference images
  • Projects serving embeddings, rerankers or multimodal models
  • Deployment workflows that need a verified container URI before endpoint creation

Common use cases

  • Choose a vLLM image for a text-generation model
  • Select TEI for embeddings or compatible rerankers
  • Choose the right serving family for multimodal or diffusion models
  • Resolve a SageMaker container URI before deployment

How does Serving Image Selection work?

The skill inspects the model task and compatibility requirements, maps the workload to the appropriate serving family and resolves the corresponding AWS image. It prioritizes current Hugging Face-curated containers where compatible and documents fallback paths when the preferred image is unavailable.

Key benefits

  • Reduces deployment failures caused by incompatible containers
  • Covers several model-task families
  • Makes model-to-image selection explicit
  • Connects directly to later SageMaker deployment workflows

Things to know

  • Container recommendations can change as serving stacks evolve
  • A compatible image does not guarantee the model fits the target hardware
  • Region and version availability can affect the final image choice

Compatible tools

Claude CodeOpenAI CodexGemini CLICursor

Frequently asked questions

Which container should I use for a Hugging Face LLM on SageMaker?
The skill maps standard text-generation LLMs to the Hugging Face vLLM serving family when compatible, with separate guidance for multimodal, embedding and other model types.
Is TEI used for every reranker?
No. The source guidance distinguishes encoder or cross-encoder rerankers from generative causal-LM rerankers, which may require different serving families.
← Back to Skills Directory