This is EnCodecMAE, an audio feature extractor pretrained with masked language modelling to predict discrete targets generated by EnCodec, a neural audio codec. For more details about the architecture and pretraining procedure, read the paper.

cd encodecmae
pip install -e .