MultimodalTransformer constructor

MultimodalTransformer({
  1. required AudioTransformer audioModel,
  2. required VideoTransformer videoModel,
  3. required int numClasses,
})

Implementation

MultimodalTransformer({
  required this.audioModel,
  required this.videoModel,
  required this.numClasses,
}) : fusionHead = Layer(
       audioModel.embedSize + videoModel.embedSize, // 64 + 128 = 192
       numClasses,
       useGelu: false,
     );