Closed I8dNLo closed 10 months ago
It could also be helpful to provide better instructions for learning with my own dataset. Like the structure of files, structure of key files and setups of env
Can you please specify the way to reproduce training on vctk dataset? I've downsampled all the .wavs and even changed naming: `p362_074_mic1.wav -> p362_074.wav
But something is wrong. For example your script looks for p362_073.wav, but nothing similar exists in original dataset
As the training data has been copied over from vits1, I assume it has been used on an older version of the vctk dataset and the newer one has a handful of files missing (~500).
I'm doing some training on vctk myself and had to clean those up, but didn't want to commit them because it might not be compatible with the new vctk dataset.
Can you put it in some other branch for example? The current instruction is misleading and only confuses potential users
I made #48, there you can check my 3 files.
You can just download them and have your config.json point to those instead of the original, that's how I do it at least.
merged in latest patch
Can you please specify the way to reproduce training on vctk dataset? I've downsampled all the .wavs and even changed naming: `p362_074_mic1.wav -> p362_074.wav
But something is wrong. For example your script looks for p362_073.wav, but nothing similar exists in original dataset