Skip to content

Latest commit

 

History

History

SageMaker

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
from_athena.py
from_athena.py
 
 

README.md

AWS SageMaker

Jump to

Useful Libs and Tools

SageMaker Studio

  • Setup
  • Data processing
    • SageMaker Data Wrangler
    • SageMaker Feature Store (Offline or Online)
      • Train models with Offiine
      • Performs low-latency inferecing with Online
      • There are three main ways to store features in Amazon SageMaker:
        1. Using Amazon SageMaker Feature Store as an Amazon SageMaker Data Wrangler destination after preprocessing steps have been completed and features have been added.
        2. Exporting a notebook from SageMaker Data Wrangler that runs through feature definition, feature group creation, and ingestion of data into SageMaker Feature Store.
        3. Using the SageMaker Python SDK in a custom notebook that runs through feature definition, feature group creation, and ingestion of data into SageMaker Feature Store.
  • Model development
    • SageMaker Experiments (similar to MLflow)
      • Use Amazon SageMaker built-in algorithms or pretrained models (link)
    • SageMaker fully-managed MLflow
    • SageMaker Debugger
    • SageMaker Operators for Kubernetes (link)
    • SageMaker Estimator - to run a training job
    • Hyperparameter tuning (link)
    • SageMaker Autopilot
    • SageMaker Clarify - to detect bias in pre-training data and post-training models and access explainability reports.
    • SageMaker JumpStart
      • Supported foundation models 220+
  • Deployment and Inference
    • Model registry
    • SageMaker Pipelines - SageMaker Model Building Pipelines steps
    • SageMaker hosting services
    • Production endpoint testing strategies
    • Model Cards - when they want to publish their model in public
  • Monitoring
    • Unit test for data - deequ
    • Schema for Violations (constraint_violations.json file) - (link)

Neuron