Request: release the audio tokenizer encoder (audio → RVQ tokens) to enable fine-tuning

Thank you for the open-weights release — the quality and the architecture write-up are both impressive.

The release currently ships only the decode-side components: the Global/Local LLMs, RVQ depth decoder, flow-matching module, Flow-VAE/vocoder (and dav.pth, which does include its codec encoder). However, the 8-layer RVQ music tokenizer's encoder — the component that maps a waveform to the semantic (layer-1) and acoustic (layers 2–8) token sequences — does not appear to be included.

Without it, fine-tuning the Global LLM is not possible, because there is no way to produce target semantic-token sequences from one's own audio data. Our use case is research on melody/harmony-conditioned generation: fine-tuning the Global LLM on paired symbolic+audio data while keeping the entire synthesis stack frozen.

Questions:

1. Are there plans to release the tokenizer encoder weights (and preprocessing code), similar to how other open music/speech releases have shipped their codecs?
2. If not, could you share the tokenizer's training configuration so the community can attempt a faithful reimplementation?

Thanks again for pushing open music generation forward.

Start by reviewing the repository's release artifacts and documentation, including dav.pth, and verify which RVQ tokenizer components are present. Done would require the 8-layer audio tokenizer encoder with preprocessing code, or the tokenizer training configuration needed for a faithful reimplementation.