These datasets are used to train lambda-audio models.