CLI version 11.4.1
Release Date: 22 March 2023
TruEra release notes are structured by product functional area and further organized as follows:
|
Product Area: DIAGNOSTICS
NEW
-
Python SDK
- Added support for creating a split in TrueraWorkspace for production monitoring.
-
Platform Integrations
- Added support for AWS-IAM Role-based Credentials
ENHANCEMENTS
-
Python SDK
- Default sample_count for add_data_split increased to 50K for remote_tru. In addition, a warning message is issued when ingestion done via data layer exceeds sample_count; message includes instructions for overriding the existing threshold.
- Error INFs now computed with SHAP.
- Kernel SHAP background capped at 100.
- Argument (arg) name for uploading error influences changed in upload_project() method to upload_error_influences from upload_error_qiis.
- To better align with Web App, suggest_high_error_segments() method changed to find_hotspots().
- add_data() method now supports production monitoring data.
- Additional check now ensures that the id of a column name is not in the feature map.
- Saved interesting segments now exposed by find_hotspots().
-
Platform Integrations
- Improved UX for SageMaker Monitoring.
-
Security
- Removal of Python vulnerabilities flagged by Trivy scan.
-
Documentation
- Code examples added in Python SDK Technical Reference for the following:
- Remaining TabularExplainer function docstrings (i.e., those that previously lacked examples)
- authentication
- compute_performance explainer.
- Doc navigation structured for better clarity and improved UX
- Code examples added in Python SDK Technical Reference for the following:
FIXES
-
Python SDK
- score_type field in add_model_predictions now uploads other scores for "classification" projects
Comments
0 comments
Please sign in to leave a comment.