source: hugging face blog: from hugging face to amazon sagemaker studio in one click
level: technical
hugging face and amazon sagemaker ai have introduced a direct integration that simplifies moving from model discovery to experimentation. developers browsing supported models on hugging face can now click customize on sagemaker ai or deploy on sagemaker ai to land directly in a pre-configured sagemaker studio environment. the selected model is pre-loaded, and the necessary permissions are automatically set up, removing the need to manually create domains, configure iam roles, or request gpu quotas.
the integration includes three main features. deep links from hugging face model pages open the model customization or deployment pages in sagemaker studio with the model context preserved. new studio environments come with a managed policy that grants permissions for fine-tuning methods like supervised fine-tuning and direct preference optimization, as well as deployment to sagemaker ai or amazon bedrock endpoints. additionally, the studio ui now shows gpu quota availability when selecting instance types, so developers can see which instances are usable without checking service quotas separately.
to use the feature, a developer selects a supported model on hugging face, clicks the relevant button, and signs into aws if needed. they then arrive on the customization or deployment page in sagemaker studio, where they can configure parameters and submit jobs. after deployment, they can test the endpoint directly within studio. this flow aims to reduce context switching and setup time, letting developers focus on experimentation and deployment.
why it matters: this integration speeds up the workflow for ai developers by cutting out manual setup steps, making it faster to fine-tune and deploy open models in a controlled cloud environment.
source: hugging face blog: from hugging face to amazon sagemaker studio in one click