Skip to content

Release Notes V1.17.0

Version 1.17.0
Release Date 05th January 2022

Changes

Data and Model Management

  • Ability to copy transforms within and across models in Model Studio for easier reusability across models

  • The model input distribution report has been made faster by leveraging spark’s parallelism more efficiently - This is noticeable when a large number of inputs have been used in a model

  • Increased flexibility in using Explainer models- users allowed to use the model function in the explainer logic and also H2O’s contribution logic can be used with MOJO files.

  • When using H2O models, users allowed to select desired output - probability values, labels, calibrated probabilities, etc.

  • Allow adding descriptions for columns in Table Registry - this allows users to add column description

  • Support sklearn pickles directly when used in model Studio without any backend configurations. This uses the new feature_names_in_ API in sklearn 1.0+ to match the feature names

Policy Management
  • Enable search for data-elements/features/models with both alias and names in the rule builder
  • Better filters for Global Variables to allow users to filter and find them easily
Common

  • Simplified sharing an object with multiple users. Earlier each user had to be added one by one - now users can select multiple users and add them together

  • Show simulation details like date filer, sampling criteria, scheduling, etc. even when the simulation is running

  • An approval queue for Global Functions has been added for approvers to easily find the functions that need to be reviewed.

  • Workspaces:

    • This is a new concept added to enable multi-tenancy. It allows different teams to have their own independent workspace in Corridor to build analytical use cases.

    • Workspaces have an independent set of registries, settings, etc. but use the same set of backend configurations (like database, authentication, filesystem, etc.)

    • To use workspaces, ENABLE_WORKSPACE configuration needs to be set to True, and the CLI command “add-workspace” can be used to add workspaces.

  • Role Management - allowed creating superusers with Write access - these users will be able to write to Draft and Pending Approval objects even if the object has not been shared with them for Write access.

Technical

  • We now support Microsoft SQL Server 2012+ as the metadata database

  • Add ability to import/export the entire database using the CLI. This allows easier backup/restore mechanisms for the database and interoperability across metadata databases.

  • Handle warnings related to is/= usage when using Corridor with python 3.8

  • Deep links for internal components like tabs inside a registry (Details, Dashboard, Jobs, etc.) have been created. This allows for easier navigation through the platform and also deep linking to access an inner section directly via bookmarks or link sharing.

  • Improve speed of report preparation in jobs - Caching for the report metadata has been enabled, which avoid repetitive fetches to the database

  • Improve speed of accessing and navigating across registry pages when users are accessing it by lazy load dropdowns and other components which are not visible directly

  • Performance of the Lineage being shown on the registry pages has been improved significantly by using graph algorithms by using graph algorithms

  • Artifact creation when there are a large number of tables (100s or 1000s of tables) registered in Table Registry has been improved. This impacts Simulations, Notebook create_data() usage, and Export artifact.