Compatibility Across Application, Flow, and Model Versions

In previous versions of Hyperscience, if you wanted to retrain a model after upgrading to a new version of the application, you needed to upgrade the model’s flow, along with all of the flow’s other models. Depending on the flow, these models could include models for Classification, Identification, and Transcription.

Beginning in v40, you can retrain a model on the latest application version and deploy it to a flow from a previous version. The “N-2 compatibility rule” still applies, meaning that you can use any model with any flow, as long as they were created in the current version of the application, the previous version, or the version preceding that version. The difference in v40 and later is that there is no longer a tight coupling between flow and model versions.    

The table below describes the compatibility of application, flow, and model versions in recent and upcoming versions of Hyperscience.

Note that the “N-2 compatibility rule” does not apply to Text Classification models, as their versions still need to match the versions of the flows that use them.

This extended compatibility of models reduces the effort required to retrain models created in previous versions of Hyperscience. If you had been proactively upgrading your models and flows during application upgrades so that you could retrain models more quickly, you no longer need to do so in v40 and later.  

Managing previous model versions

You can view a list of the available versions of a model in the Model History card on the model’s details page.

From this card, you can export any available version of the model and you can deploy the version of the model that best meets your needs. Only models that are compatible with your application version are listed.