Property EnableModalityAugmentation
- Namespace
- LMKit.TextAnalysis.Training
- Assembly
- LM-Kit.NET.dll
EnableModalityAugmentation
Gets or sets whether to add modality-augmented samples when the engine runs in Multimodal.
public bool EnableModalityAugmentation { get; set; }
Property Value
Examples
using var model = new LM("path/to/model.gguf");
using var pii = new PiiExtraction(model);
var dataset = new PiiExtractionTrainingDataset(pii)
{
// Enable automatic augmentation for multimodal samples
EnableModalityAugmentation = true
};
// This single call produces 3 samples when engine is Multimodal:
// one Multimodal, one Text-only, and one Vision-only
dataset.AddSample(
"Patient ID: P-12345",
new[] { new EntityAnnotation("PatientID", "P-12345") });
dataset.ExportAsSharegpt("augmented_dataset.json", overwrite: true);
Remarks
When true and the generated sample uses Multimodal, two additional samples are appended automatically using Text and Vision with the same content and labels. This can improve robustness across modalities during fine-tuning.