ADS 2.13.2

• Fixed bug in DataScienceModel defined metadata.

ADS 2.13.1

ADS 2.13.1

• Upgraded OCI SDK to v2.148.0 for improved compatibility and performance.
• Added support to launch the AI Quick Actions API server from anywhere using:

      python -m ads.aqua.server

• Introduced HttpxOCIAuth class, enabling HTTPX clients to sign requests using OCI signers, ensuring compatibility with OCI Model Deployment services and LLM models.

• Forecast Operator:

  • Optimized the auto-select feature to prevent redundant data re-saving during cross-validation.
  • Implemented AutoMLx Global Explainer for enhanced model interpretability.

• Bug Fixes:

  • Fixed the Python 3.11 Compatibility. Replaced the deprecated cgi.parse_header function with a custom pure-Python implementation for parsing Content-Disposition headers.
  • Resolved a recursion issue that occurred when creating a new model with the latest OCI SDK version.

v2.13

• Bug fix related to latest OCI SDK version compatibility

ADS 2.12.11

• Fixed bug while loading model configuration in AI Quick Actions.

ADS 2.12.10

• Support inference container URI to be edited for unverified models in AI Quick Actions.
• Logging improvements for all CRUD operations in AI Quick Actions.
• Expand allowable parameters for fine-tuning operation in AI Quick Actions.
• Allow local model cache delete while registering models in AI Quick Actions.
• Add validation flag ignore-model-artifact-check to ignore config.json check while registering model in AI Quick Actions.
• Allow user to watch fine-tuning job run logs in AI Quick Actions when using CLI mode.
• Load base model configuration files by default for verified models in AI Quick Actions.
• Added a framework to support embedding onnx models.
• Added headers parameters for ADS Langchain.
• Addition of SessionLogger and MetricLogger AutoGen loggers for logging and reporting.
• Support for a centralized and reusable client for interacting with OCI Model Deployment.
• Bug fix for ADS telemetry logging.
• Set defaults for optional parameters in ChatOCIModelDeploymentVLLM.
• Real-Time What-If Analysis with Model Catalog and Model Deployment Integration.
• Support for AutoMLx internal explainability mode.
• Standardize outputs & report for single-series forecasts without target_category_columns.

ADS 2.12.9

• Support freeform and defined tags for resource creation in AI Quick Actions.
• Enhance LangChain Integration to accept the route parameter in the request headers.

ADS 2.12.8

• Fixed bugs in the AI Quick Actions Model Deployment for models registered via Object Storage.

ADS 2.12.7

• Fixed bugs in the AI Quick Actions Evaluation.

ADS 2.12.6

• Support using environment variable to override AI Quick Actions Fine Tuning container image.
• Adding metadata in telemetry error logging for AI Quick Actions.

ADS 2.12.5

• Fixed bugs in the AI Quick Actions Evaluation.