Training with augmentations question

Where can I read deeply on how is ’ Training with augmentation’ performed? Let’s say I have 100 files in my training set?
If I apply

  1. Standard deviation for Gaussian additive noise: --data_aug_features_additive
  1. Standard deviation for Normal distribution around 1 for multiplicative noise: --data_aug_features_multiplicative

Will my model only train on 100 original files + 200(100 per each augmentation) or only on 100 augmented files with two augmentations? What is the idea behind this?

The best source of truth will be the source code :slight_smile: