ADS 2.13.11

• AI Quick Actions: Added support for LlamaCpp in the client to enable the new predictWithResponseStream endpoint for streaming inference.

ADS 2.13.10

• AI Quick Actions: Enhanced the client to support the new predictWithResponseStream endpoint for streaming inference.

ADS 2.13.9

• Additional precision support in ForecastOperator.
• AI Quick Actions: Use defined-metadata to include configuration for fine-tuned models.
• AI Quick Actions: Support for embedding models in a multi model deployment.
• AI Quick Actions: Fixed a bug in multi-model deployment to use model artifact json directly instead of accessing service bucket when creating a new grouped model.
• AI Quick Actions telemetry improvements and enhancement to use threadpool instead of creating unbounded number of threads for telemetry.
• AI Quick Actions: Support for list API for compute capacity reservations to onboard Bring-your-own-reservations (BYOR).
• AI Quick Actions: Fixed a bug which now allows multiple parameters for deployment parameters.
• AI Quick Actions: Enhances the model deployment logic for vLLM architecture version.
• AI Quick Actions: Enhances functionality to retrieve deployment configuration for fine-tuned models.

ADS 2.13.8

• AI Quick Actions: Adds support for image-text-to-text models in multi-model deployment.

ADS 2.13.7

• AI Quick Actions: Removing OSS bucket config dependency.
• Fixed a bug in the ADSDataset type inference logic. Resolved an issue where the dataset incorrectly identified column types during automatic type discovery.

ADS 2.13.6

• Fixed PII operator report generator script to use the latest code from report-creator.
• AI Quick Actions: Enhance the container family validation for multi-model deployment.

ADS 2.13.5

• Added support for OpenAI and AsyncOpenAI clients to enable seamless integration with OCI-based LLM model deployments.
• AI Quick Actions: Enhanced error message handling for clearer diagnostics and improved user experience.
• AI Quick Actions: Introduced support for Multi-Model Deployment, allowing users to deploy multiple models within a single endpoint.

ADS 2.13.4

• Added compatibility with Python 3.12.
• Forecast Operator: Introduced auto-select fallback to AutoMLX for faster approximate mode execution.

v2.13.3

What's Changed

• Forecast Operator
  ** Enabled custom report title
  ** Added AutoMLx to model list
  ** Added date to explanation csv
• Fixed defined tag permission issue when creating AI Quick Action fine-tuning and evaluation.

ADS 2.13.2

• Fixed bug in DataScienceModel defined metadata.