This github repository contains the code leading to the paper:
Automatic Data Augmentation Selection and Parametrization in Contrastive Self-Supervised Speech Representation Learning by Salah Zaiem, Titouan Parcollet and Slim Essid.
The repository contains two folders for the two main steps. The folder selection contains the code needed to compute the scoring of a given augmentation distribution. The folder "experiments" contains the code needed to perform the cola pretraining and the Speaker ID and language ID downstream finetunings. The snippets to run every step are available within the folders.