AIF-C01 Practice Question 42: Pretrained Open Models
AWS Certified AI Practitioner (AIF-C01) · Test 5 · question 2 of 10
A data science team wants to start from a publicly available, pretrained open-source model instead of training one from scratch. The team wants to browse a catalog of such models, fine-tune and evaluate one on its own data, and deploy it to an endpoint in its own AWS account with only a few steps in Amazon SageMaker Studio. Which AWS capability should it use?
Show answer and explanation
Answer:
Amazon SageMaker JumpStart provides pretrained, open-source models for a wide range of problem types. From the models landing page in SageMaker Studio, you can deploy, fine-tune and evaluate models from popular model hubs, and JumpStart also offers solution templates and example notebooks. AWS names it the low-code way to deploy pre-trained models to preconfigured inference endpoints (option A). As with any third-party model, the team is responsible for reviewing its license terms.
- B is wrong because Model Cards *document* a model's intended uses, risk rating and evaluation details for governance. They don't supply pretrained models. - C is wrong because Guardrails filter harmful content and sensitive information in the inputs and outputs of generative AI applications. They aren't a model catalog. - D is wrong because Prompt Management stores, versions and reuses *prompts*. It doesn't provide models to fine-tune or deploy.
Reference: SageMaker JumpStart pretrained models