Kwisss
Kwisss
> but the api keeps responding with only släktet which mean "the genus" in Swedish. So it's clearly nothing to do with seed Are u swedish? do you by any...
Sorry I hadn't read your response, I think I got it working now. My understanding now If I mask the noise in inject noise, I need to mask the conditioning...
> > 模型 > > > 参数量小的模型是这样的，底模很容易被洗干净，试听差异不是很明显，可以推理一下试试看 > > 目前练出来的48Ｋ感觉和目标原音也不太像，是不是因为我的数据集不是歌声的原因 想试试16Ｋ的又一报错 ValueError: num_samples should be a positive integer value, but got num_samples=0 有人说路径不对，但查了下并没有发现哪不对 indeed, I had this problem when I...
Yeah I have the same problem.
> Most of the potential performance is not yet available. T3 inference uses the 0.5B Llama 3 via _transformers_ which is where the performance issue lies. It has a lot...
> Hey [@Kwisss](https://github.com/Kwisss) really like your performance statistics, is it possible for you to share your fork with these performance improvements? Yes i can @mushitori here it is: The code...
the gguf model also throws this error
> > > > > which for the 10M context length is going to be giant > > > > > > > > > > > > I don't...