AWS On Air ft. Amazon SageMaker Inference Recommender
summary
Amazon SageMaker Inference Recommender automates load testing and tuning to pick the best deployment: it benchmarks a model package (demo uses ResNet-50 with image payload) across 70+ EC2 ML instance types, runs a 30–60 minute job, tries compiled artifacts, and returns instance type, tuned server/kernel params, initial scaling values, and metrics (invocations, model latency, cost per inference). It accepts Deep Learning Containers or custom containers, integrates with Model Registry, and outputs concrete recommendations for production.
AI-written summary · about
chapters (10)
- 0:00Welcome and introduction
- 1:13Introducing SageMaker Inference Recommender
- 3:23Target customers and use cases
- 6:33Supported containers and payloads
- 8:21Demo overview and setup
- 12:17Default vs advanced experience
- 14:03Demo packaging the model
- 16:11Running the recommendation job
- 17:33Interpreting results
- 19:42Resources and wrap-up
related talks
48mAmazon SageMaker — Survey of Ways to (Re)deploy Models on AWS
19mUsing Amazon SageMaker to develop and deploy ML models
01h 00mHigh-performance & cost-effective model deployment with Amazon SageMaker
19mDeveloping large-scale machine learning algorithms on Amazon SageMaker
32mOptimizing AI/ML workloads for sustainability
41mDeploy ML models for inference at high performance & low cost
44mPragmatic Machine Learning in the Cloud
35mAccelerate deep learning in the cloud with custom ML environments