YuE2 GGUF files by Serveurperso, kept unchanged

Unchanged copies of seven files published by Serveurperso in Serveurperso/YuE2-GGUF (revision e630f2b8), the GGUF conversions that yue2.cpp runs. They are kept here so that Yeufonic's installers still work if the original moves or is taken down. The installers fetch from here first and from the author's repository if this copy cannot be reached; either way the file must match the checksum below.

All credit for these files belongs to their author, and for the models to their original authors (below). Please use, cite and follow the original repository.

File Bytes SHA-256
YuE2-3B-Q8_0.gguf 3,810,232,064 41121ce97786d7795a325bcf123ca196956c03bb252c9e75384cfc1f2fc19e6b
YuE2-Vae-F32.gguf 530,497,344 93e49dfb1970e89ad64cacb17cf13b5d05f6bb30ef7ed3adae3050bcb728638a
SheetSage2-Q8_0.gguf 105,629,856 95d8fa8761753b4833782a52b68d5c0afa3bd3cfbc03d3dcbbf8ded0c13f611e
MERT-v2-FullSong-Q8_0.gguf 902,283,712 35a6032ec7e8123c02becdefc7bfe10a3adb33faa8a1951dc0af864b4494b8ef
YuE2-3B-Q6_K.gguf 2,943,331,072 42068d5e57713df11b9f6a5a3751072d9bbefe231d1d088f9c100e026ffa4ef9
YuE2-3B-Q5_K_M.gguf 2,622,936,832 cd3efd250b734a229172800f08f33a893ad1f66b521666e176cdae4a0729b281
YuE2-3B-BF16.gguf 7,166,072,352 668ca9ffa4622449916aae5068b5737363e7ec186a824cd3d7a1a8a95269f166

What each one is for, in Yeufonic's yue2.cpp engine:

  • YuE2-3B-Q8_0.gguf: YuE2 3B backbone, Q8_0: plans and renders (the size Yeufonic installs).
  • YuE2-Vae-F32.gguf: YuE2 audio VAE: renders.
  • SheetSage2-Q8_0.gguf: SheetSage2 transcriber head: a recording to a score.
  • MERT-v2-FullSong-Q8_0.gguf: MERT-v2-FullSong audio encoder: the transcriber's encoder.
  • YuE2-3B-Q6_K.gguf: YuE2 3B backbone, Q6_K: a smaller size that can be chosen.
  • YuE2-3B-Q5_K_M.gguf: YuE2 3B backbone, Q5_K_M: the smallest size that can be chosen.
  • YuE2-3B-BF16.gguf: YuE2 3B backbone, BF16: full precision, can be chosen.

Licence

CC BY-NC 4.0, as the original repository declares. Non-commercial use only, with attribution. LICENSE is the original repository's licence file, unchanged. As it says, the original work is YuE2 by MAP (m-a-p/YuE2-3B, m-a-p/YuE2-Vae); the transcriber is SheetSage2 by MAP (m-a-p/SheetSage2) on MERT-v2-FullSong by MAP (m-a-p/MERT-v2-FullSong). The conversion to GGUF and the quantization are Serveurperso's; nothing here is changed. manifest.json lists every file with its size, checksum and the revision it was copied from.

Downloads last month
251
GGUF
Model size
0.6B params
Architecture
mert2
Hardware compatibility
Log In to add your hardware

5-bit

6-bit

8-bit

16-bit

32-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support