03.12.2021
New Releases, Enhancements, Changes + Arize in the News!
Various Environments for a Single Model Version
Benefits of Environments:
- Easily analyze if there is drift in features/predictions/actuals between production and a pre-production environment such as training validation
- Compare model performance metrics from training/validation to production
- Training/validation windows for a model are more prominently surfaced to help save you time when comparing against production.
How to Use:
A Performance tab is now available under the Model Page for Score & Regression Model Types. It will come soon for Binary and Multi-Classification Models - stay tuned!
This tab surfaces up key performance metrics specific to the model type. For score models, we have dynamic AUC-ROC analysis:
Score Model Performance Tab
We added the ability to quickly see which alerts are triggered/healthy without having to directly visit the Monitors Page. You can now see a consolidated view of a model's health from the Model Page.
New Version of Python SDK
We released a new version of the Python SDK, which requires the Model Version and a Model Type on Score Categorical Models only. For more details, check out the Python SDK Changelog here.
Arize is a proud partner of the Tecton Apply ( ) Conference! The conference has a great line up of ML practitioners speaking about how they are getting ML into the real world. Register to attend!
Last modified 2yr ago