This is an outstanding work on music generation. It feels like a real "GPT moment" for music AI. Or more accurately, a "DeepSeek moment" since your team make an open-source breakthrough.
I went through the paper and played with the model, and the quality honestly caught me off guard. The music stays coherent over long spans and actually sounds musical, not just technically correct.
The code is also super tidy and well thought out. Definitely one of those repos you learn a lot from just by reading it.
Massive respect to the authors from PKU and many Top-tier colleges, and thanks for putting this out there ❤️❤️