MENU
      Contents x
      models.install
      • 11 Nov 2020
      • 1 Minute to read
      • Contributors
      • Print
      • Dark
      • PDF
      Contents

      models.install

      • Updated on 11 Nov 2020
      • 1 Minute to read
      • Contributors
      • Print
      • Dark
      • PDF
      Article summary

      Description

      Initiates an asynchronous installation of a specified machine learning (ML) model from your app's locker.

      When installation completes, the model is stored inside the runtime's cache object using a provided key.

      The Ux.ResourceReady event is triggered so you can take action when your model becomes available.

      Input Parameters

      NameTypeRequiredDescription
      tokenTextYesModel's locker token
      keyTextYesUnique cache key where model will be stored

      Returns

      No return value.

      Examples

      // inside the Ux.Initialize script...
// install model from app's locker using its token
ai.models.install("TOKEN", "riskModel");
      JavaScript

      Model Formats

      The following model formats are supported:

      NameDescription
      javascriptJavascript model format
      bayesialabBayesialab model format
      tf *Tensor Flow model format
      onnx *Open ONNX AI CNN format

      Formats with an asterisk coming soon

      Remarks

      Use this function to install a previously trained machine learning (ML) model that you would like to use while your app is running.

      One common scenario is to install a model during intialization. Next, as a user interacts with your app in realtime, gather their inputs (using the Ux.Change event) and run them through your model. Finally, display model results inside the attached insights panel or adjust your app's language output.

      Two-Step Process

      Installing models can be slow based on connection speeds and the size of the models being downloaded.

      As a result, downloading models requires a two-step process. First, you must initiate the installation from one script (typically Ux.Initialize). Second, you run the model (using ai.models.run) after the installation completes from another script (typically Ux.ResourceReady or Ux.Change).

      By separating these steps, the user's experience remains fluid. Otherwise, the user interface would freeze until the installation completes, resulting in a poor user experience.

      What's Next