How oracle/accelerated-data-science works
Main Components
a. Core SDK Layer:
• Core modules: common, model, dataset, feature_engineering
• Connection to OCI authentication and configuration
b. Service Integration Layer:
• OCI services that ADS interfaces with:
• Data Science service
• Model Catalog
• Model Deployment
• Jobs
• ML Pipelines
• Data Flow
• Object Storage
• Vault
• Big Data Service
• Data Catalog
• Autonomous Database
c. Feature Modules Layer:
• Functional modules:
• Data Loading (database, oracledb, mysqldb)
• Model Management (model, deployment)
• Feature Engineering (feature_engineering, type_discovery)
• ML Operations (jobs, pipeline)
• Evaluation & Tuning (evaluations, hpo)
• LLM Integration (llm, aqua)
d. Extensions and Tools Layer:
• CLI tools (opctl)
• Operators (forecast, anomaly, etc.)
• Visualization components
Relationships and Data Flow
• Authentication flow from applications through the SDK to OCI services
• Data flow between components
• Model lifecycle flow
• Job execution flow
Implementation Details
• Component Name |
• Key Features |
• Supported Formats |
• Integration Points |
[Client Applications/Notebooks]
• ↓
[ADS Core (Authentication/Configuration)]
• ↓
[Feature Modules] ↔ [OCI Services]
• ↓
[Extensions/Tools/CLI]
Technical Details to Highlight
• How the SDK handles different authentication methods
• Model lifecycle management
• Demonstrate the job execution flow
• Data ingestion patterns
• Integration points with ML frameworks