Dan Lyth

Not solved for me either

I'm seeing the same thing. Did you find a fix? Also, are you able to pick up training from where you left off? I'm on CUDA 10.1, Windows 10, Pytorch...

Nice, that's a good solution. And yeah, I found out this morning that it picks up from where it left off very well. How many epochs did you leave it...

Thanks. I was going to try fine-tuning but I have a 17.5 hour dataset so thought I would just train from scratch as it's not too much smaller than the...

Just a wee note to say that yes, it can be repurposed as a TTS model. I used Audiocraft as my codebase for this paper: https://arxiv.org/abs/2402.01912

I'm interested in this too. In [gen_unprompted_outputs](https://github.com/facebookresearch/audiocraft/blob/a2b96756956846e194c9255d0cdadc2b47c93f1b/audiocraft/solvers/musicgen.py#L535), the `prompt_duration` is set to [prompt_duration = dataset_duration / 4](https://github.com/facebookresearch/audiocraft/blob/a2b96756956846e194c9255d0cdadc2b47c93f1b/audiocraft/solvers/musicgen.py#L489) rather than `None`. If I understand this correctly, "unprompted" samples are still being...

Sorry I missed this. Yeah, a combination of both reasons; the HuBERT model and k-means model are available, and also, a bunch of literature prior to AudioLM has used this...

You need to be on Python 3.10.7, `pyenv` is your friend.