使用 Amazon SageMaker、Amazon Bedrock 和 AWS IoT Greengrass 进行 AI 模型基准测试.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Ben SnyderSr.Applied ScientistAmazon Web ServicesAnshul GangSr.Global TPM Specialist and ArchitectureAmazon Web ServicesAI model benchmarking with Amazon SageMaker,A

2、mazon Bedrock&AWS IoT GreengrassA R C 3 3 4 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AI for Real Time Decision making at the Edge 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Small models are needed for real time inference,but lack flexibilityLarge mod

3、els better understand images,but cant run in real time 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Model distillation uses larger models to teach smaller models 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Edge Inference 2025,Amazon Web Services,Inc.or it

4、s affiliates.All rights reserved.Edge Hardware:Hailo AI Accelerator 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Model Selection:YoloV11 at the edge Claude for agentic descriptions LLMDet for bounding box labelling 2025,Amazon Web Services,Inc.or its affiliates.All rights reser

5、ved.YoloV11 Non-max supression 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AWS Services 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AWS Bedrock and Strands Agents 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.SageMaker Training Jobs

6、2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Device Deployment 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.R E S O U R C E SA W S S A G E M A K E R A IA W S B E D R O C KS T R A N D S A G E N T S