Arize AI

Model Version

Model Version captures a minor or major change to the model where the model could evaluate differently given different input data.


Model Version captures a minor or major change to the model where the model will evaluate differently given different input data.
[Recommended] A new model version should be created when:
  • A new model is trained -- there are a new set of weights
  • A new feature is added
Example: Model A has 3 versions - v1, v2, and v3

Code Example

Each model version is defined in the code when the data is sent into the platform:
response = arize.log(
model_version='v1' .....)

Examples of Model Versions across Arize

Model version selectors are available extensively in the platform. You can select ONE version, MULTIPLE versions, or across ALL versions.
In Monitors, you can select specific versions or monitor across all versions.
Versions in Monitors
In the model's Performance tab, you can choose to view performance for a specific model version, or even compare performance of different model versions to see which version performs better.
Version in Model Performance Tab
The Explainability tab supports model version analysis. The SHAP values can be compared across model versions to understand the impact of various features on a model's predictions.
The majority of widgets in the Arize platform have version selectors, allowing you to deeply analyze and compare the performance of different model versions. This is especially helpful in A/B testing improvements/updates to your models.
Versions in Dashboard Widgets
The dashboards have version filters that are available to selectively filter the data based on a specific model and version.
Dashboard Filter Model Version
Questions? Email us at [email protected] or Slack us in the #arize-support channel