[MLS-C01] [IMPL and OPs] Build solutions

Build machine learning solutions for performance, availability, scalability, resiliency, and fault tolerance

Posted by Oscaner on August 7, 2022


  • SageMaker makes it simple to develop high quality models by enabling quick deployment at scale
  • SageMaker provides the tools needed for machine learning in a single toolset that allows you to get models to production faster with much less effort and at lower cost than with the traditional tool suite
    • SageMaker Studio
    • SageMaker Autopilot
    • SageMaker Experiments
    • SageMaker Debugger
    • SageMaker Model Monitor

Build with SageMaker Studio

Single, web-based visual interface where you can perform all your machine learning development steps

  • Build, train, and deploy models
  • Upload data, create new notebooks, train and tune models, move back and forth between steps to adjust experiments, compare results, and deploy models to production

Performance with SageMaker Autopilot

Build, train, and tune models with complete visibility and control

  • Inspects input data, applies feature engineering, picks optimum set of algorithms, trains and tunes multiple models, tracks model’s performance, and ranks your models based on their performance
  • Gives you the best model for your problem at hand while saving time
  • Can be used by developers without extensive machine learning experience

Evaluate with SageMaker Experiments

Organize your artifacts, track metrics, and evaluate training runs using SageMaker Experiments

  • Manage your model iterations by capturing the input parameters, configurations, and results, and saving them as experiments

Fault Tolerance with SageMaker

Analyze, identify, and alert problems for machine learning

  • Track real-time metrics when training such as validation, confusion matrices, and learning gradients to increase mode accuracy
  • Generate warnings and corrective action advice when you experience training issues

Resiliency with SageMaker Monitor

Detect and remediate concept drift where the patterns used to train the model have changed over time

  • Identifies concept drift in deployed models and gives detailed notifications that help identify the source of the drift

本文由 Oscaner 创作, 采用 知识共享署名4.0 国际许可协议进行许可
本站文章除注明转载/出处外, 均为本站原创或翻译, 转载前请务必署名