Notebook1: Training and Hosting a PyTorch Model Notebook2: How to Use Spot Training • デモ SageMaker Debugger, SageMaker Experiments, SageMaker Feature Store MLOps • まとめ: Q&A とアンケート Agenda http://bit.ly/sage-615
for machine learning SageMaker Data Wrangler NEW Aggregate and prepare data for machine learning SageMaker Processing Built-in Python, BYO R/Spark SageMaker Feature Store NEW Store, update, retrieve, and share features SageMaker Clarify NEW Detect bias and understand model predictions BUILD SageMaker Studio Notebooks Jupyter notebooks with elastic compute and sharing Built-in and Bring your-own Algorithms Dozens of optimized algorithms or bring your own Local Mode Test and prototype on your local machine SageMaker Autopilot Automatically create machine learning models with full visibility SageMaker JumpStart NEW Pre-built solutions for common use cases TRAIN & TUNE Managed Training Distributed infrastructure management SageMaker Experiments Capture, organize, and compare every step Automatic Model Tuning Hyperparameter optimization Distributed Training NEW Training for large datasets and models SageMaker Debugger NEW Debug and profile training runs Managed Spot Training Reduce training cost by 90% DEPLOY & MANAGE Managed Deployment Fully managed, ultra low latency, high throughput Kubernetes & Kubeflow Integration Simplify Kubernetes-based machine learning Multi-Model Endpoints Reduce cost by hosting multiple models per instance SageMaker Model Monitor Maintain accuracy of deployed models SageMaker Edge Manager NEW Manage and monitor models on edge devices SageMaker Pipelines NEW Workflow orchestration and automation Amazon SageMaker SageMaker Studio Integrated development environment (IDE) for ML
Amazon S3 The Jupyter Trademark is registered with the U.S. Patent & Trademark Office. トレーニングでのメリット: • API 経由で学習⽤インスタ ンスを起動、 学習が完了すると⾃動停⽌ • ⾼性能なインスタンスを 秒課⾦で、 簡単にコスト削減 • 指定した数のインスタンス を同時起動、 分散学習も容易
Model Development Model Training & Evaluation Model Deployment & Inference Production Integration Data Engineers Data Scientists ML Engineers AWS Accounts, Controls, Dev environments, and MLOps stacks (DevOps tools, artefacts repos, ML logs insights) SysOps ML Workflow Automation - Model Management - Continuous Delivery
All rights reserved | How Amazon SageMaker Pipelines works パイプライン実行の 開始: • 手動 • データアップロード 時の CloudWatch event • コード check-in (git push) Acceptable accuracy Non-acceptable accuracy Get input data Process data Train model Validation Deploy model Alert and stop