Training Set - Interactml/iml-unreal GitHub Wiki

What is a Training Set?

A Training Set holds a collection of examples and the Expected Ouputs associated with them.

See Collecting Parameters for more on Examples 👉

See Labels for more on Expected Outputs 👉

Training Sets are assets you add to your project and are stored in separate data files.

See Asset Types and Your Data for more 👉

Training Sets are populated with the Example Recorder Blueprint node.

See Recording Examples for more 👉

Training Sets are used to train a model so it can learn to associate examples with outputs.

See Training Models for mode 👉

Creation

Training Sets can be created as Unreal assets within your project structure using the Content Browser context menu.

Training Set Asset Icon
Training Set
Examples a specific model was trained with

See Asset Types for more information on asset types 👉

Training Set View

The Training Set asset type has a custom view associated with it that can be opened by double-clicking on it.

Training Set Properties

General information about the Training Set are shown at the top.

  • Name - Asset name
  • Examples - Quantity of examples recorded in the Training Set
  • Labels - Quantity of distinct output label values used by the examples
  • Label Type - Number for simple numeric labels, or a Label Type asset name for composite labels
  • Sample Mode - Single/Series sampling mode
  • Parameters - Numerical parameter values recorded for each example sample (number of floating point values)

Example Properties

The list view shows all the recorded examples and information about each one (where available).

  • #/Number - Identifier used to refer to each example in the Training Set
  • Label - The Expected Output value (or index) associated with this example
  • Samples - Number of sample parameters recorded for this example (applies to Series sampled examples)
  • Duration - Approximate time taken to record the example (applies to Series sampled examples)
  • Session - Date and time stamp when the example was recorded (where available)
  • User - Who recorded the example (where available)
  • Notes - Unused at the moment

Editing

The training set example data can be edited using selection and the toolbar buttons. Editing operations can be un-done with the usual Unreal Editor Undo function (Ctrl+Z).

Delete

To delete only certain examples; select one or more in the list and press the Delete button on the toolbar (or press the Delete key). To select multiple examples hold Shift (range) or Ctrl (individual) while clicking.

Reset

To delete all examples; press the Reset button on the toolbar.


👈 Running Models | 🏠 Home | Labels 👉

⚠️ **GitHub.com Fallback** ⚠️