YuE2-3B is now in the audio.cpp dev branch. Q8 GGUFs provided. < 9 GB VRAM. 3min40s songs generated in 60s on RTX 5090

#3
by audio-cpp - opened

Thank you for releasing this great model! We’ve integrated it into audio.cpp so hopefully more people can access it.

Note that the model is currently in the dev branch for community testing and optimization, and it won’t be included in the prebuilt binaries before merging.

Update: tonight-awake test

Combo Audio dur Wall RTF Peak VRAM
BF16 main + F32 VAE 224.96s 60.46s 0.2688 12535 MiB
Q8_0 main + F16 VAE 194.84s 38.81s 0.1992 8867 MiB
Q4_0 main + F16 VAE 221.12s 44.15s 0.1997 7755 MiB

Dev branch: https://github.com/0xShug0/audio.cpp/tree/dev
GGUFs: https://huggingface.co/audio-cpp/Yue2-3B-GGUF

Demo uses the official longform test case.

Thank you for releasing this great model! We’ve integrated it into audio.cpp so hopefully more people can access it.

Note that the model is currently in the dev branch for community testing and optimization, and it won’t be included in the prebuilt binaries before merging.

Update: tonight-awake test

Combo Audio dur Wall RTF Peak VRAM
BF16 main + F32 VAE 224.96s 60.46s 0.2688 12535 MiB
Q8_0 main + F16 VAE 194.84s 38.81s 0.1992 8867 MiB
Q4_0 main + F16 VAE 221.12s 44.15s 0.1997 7755 MiB

Dev branch: https://github.com/0xShug0/audio.cpp/tree/dev
GGUFs: https://huggingface.co/audio-cpp/Yue2-3B-GGUF

Demo uses the official longform test case.

@audio-cpp upload the build here please

@BlairRiver Building now! Once finished, you can find them in the “Artifacts” section on this page: https://github.com/0xShug0/audio.cpp/actions/runs/34642627647

@BlairRiver Building now! Once finished, you can find them in the “Artifacts” section on this page: https://github.com/0xShug0/audio.cpp/actions/runs/34642627647

Thank you very much!

This comment has been hidden (marked as Resolved)

Sign up or log in to comment