[ICLR 2025] SOTA discrete acoustic codec models with 40/75 tokens per second for audio language modeling
