SageMaker

Access:

• SageMaker notebook on EC2
• SageMaker console

Data:

• S3

Training Job

• data
• compute resources
• output
• training code

Deployment

• endpoint
• batch transform

Built-in Algorithms

• Linear Learner
  • Input: RecordIO, CSV, file or pipe
  • Normalization
  • Hyperparameters:
    • class weights
    • learning rate
    • l1 or l2/weight decay
• XGBoost
• Seq2Seq
• DeepAR
• BlazingText
• Object2Vec
• Object Detection
• Image Classification
• Semantic Segmentation
• Random Cut Forest
• Neural Topic Model
• Latent Dirichlet Allocation
• KNN
• Reinforcement Learning
• Hyperparameter Tuning

Spark and SageMaker

• sagemaker-spark
• SageMakerEstimator returns a SageMakerModel

SageMaker Studio

Machine learning IDE

SageMaker Experiments

Tracking and managing experiments

SageMaker Debugger

• Saving gradients, model states or logs for debugging;
• Reports
• Build-in rules:
  • Monitor system metrics
  • Profile model operations
  • Debug model parameters
• SMDebug client
• Insights Dashboard
• ProfilerRule
• Notifications and actions

SageMaker Autopilot

It supports:

• Algorithms

• Data preprocessing

• Model tuning

• Infrastructure

• Human in the loop

• Classification or Regression

• Tabular data

• Integrates with SageMaker Clarify

SageMaker Model Monitor

• Drift:
  • Data
  • Model
  • Bias
  • Feature
• Outliers and anomalies
• New features
• Integration with SageMaker Clarify for bias