Choose a SageMaker container image with AI
Match a model to an available SageMaker inference image. Check the serving engine, region and instance requirements before passing the image URI to deployment.
To choose a SageMaker container image with AI, describe the model and the hardware it will serve on. The hf-cloud-serving-image-selection skill checks the current AWS catalogue for the relevant image family, platform and regional address.
An embedding service and a text generator can have quite different runtime needs. Looking up the actual tag also avoids guessing a registry address from another region. Where a private network needs its own copy, the guide includes a separate ECR mirroring step.
What it does
- Find the published image family for a model type.
- Check that the entry targets SageMaker.
- Look up the correct regional registry and tag.
- Review GPU and runtime compatibility.
- Mirror an approved image to private ECR when required.
Install and use
- Install the skill. Have the model details and AWS region ready. Private mirroring additionally needs Python, Docker and an authorised AWS CLI profile.
- Ask “Find a SageMaker image for this embedding model on the proposed instance. Check the current catalogue and give me the exact URI.”
- Verify platform, region and tag. Confirm the private registry separately if the network requires a mirrored copy.
FAQ
Source
Made by Hugging Face. Original skill: hf-cloud-serving-image-selection, published on GitHub under the Apache-2.0 licence.
View on GitHub