Hi and thanks for the awesome work

Can you provide more details on the dataset and language coverage of the hubert model you trained that's used for semantic distillation of your codec?