3dluvr

Take a look at this repo: https://github.com/JonathanFly/bark instead of the suno-ai, as more work was done to it.

By default it's storing the models in ~/.cache folder, which to me is unacceptable when running in VENV. I did submit an issue about it and they came back pretty...

Yeah, I don't like running anything in the "cloud", so everything is local as much as possible.

> this is going to break the fucking internet not until voice cloning fully works, then the Armageddon will happen.

> we are looking into ways of verifying your own voice, but these things are much harder to control in a fully public releases, so might be reserved for suno...

If you pull the 0.2.1 source from GH and try to compile with CUDA, the issue is not Torch, but https://github.com/PanQiWei/AutoGPTQ/blob/main/setup.py#L13 which is `IN_GITHUB_ACTIONS = os.environ.get("GITHUB_ACTIONS", "false") == "true" `...

@TheBloke it appears from your output that it does compile the CUDA extension. What does not work is correct versioning of then compiled wheel. When installed and you do `pip...

This still needs fixing: https://github.com/PanQiWei/AutoGPTQ/blob/main/setup.py#L23 It will never mark the module VER+cuXXX unless it's being compiled through a GH action, so all local users will assume there's no CUDA support...

I think the question is whether AutoGPTQ would ever be used with non-CUDA capable cards. If CUDA only, the extension should always be built regardless of any setting, because what...

You need to specify the baudrate...for me it starts working and the bulb flashes a few times but then I get an error ``` $ python hue-thief.py -b 115200 /dev/ttyACM0...