Skip to content
Kitbay

hf-cloud-sagemaker-production-defaults

Apache-2.0Updated on Nov 11, 202611,334GitHub starsIncludes scripts

The deployment step of Hugging Face's SageMaker skills.

What the skill teaches

The deployment step of Hugging Face's SageMaker skills. Once a container image and execution role are known, it has the agent create a SageMaker model, endpoint configuration and endpoint (real-time, real-time with scale to zero, async, or an inference component) together with autoscaling, CloudWatch alarms for latency and errors, and consistent tags for later cleanup. Data capture stays off unless asked for. It insists on a smoke test rather than trusting the InService status, and covers testing, teardown and failure handling. Bundled Python scripts deploy, invoke and tear down endpoints.

Capabilities

  • Deploy SageMaker endpoints with autoscaling and alarms
  • Create async or scale-to-zero endpoints
  • Smoke-test and invoke an endpoint
  • Tear down deployed resources by tag