-
https://git.ecker.tech/ aims to provide a place to share my efforts while maintaining true ownership of my code, as I do not trust GitHub.
XMR: 4B9TQdkAkBFYrbj5ztvTx89e5LpucPeTSPzemCihdDi9EBnx7btn8RDNZTBz2zihWsjMnDkzn5As1LU6gLv3KQy8BLsZ8SG
- Joined on
2022-10-10
No I'm stupid, I had the symlink line wrong. Should be remedied in commit f31ea9d5bc95fd5f6521ef45ca7c4d5af01e1ca9.
Change the drive mount block to:
# only run once, this will save all…Should be remedied in f31ea9d5bc95fd5f6521ef45ca7c4d5af01e1ca9. I forgot to do a simple check if something exists when importing existing settings.
But again, that button's for reusing existing…
For fresh trainings, ignore the import button. The import button is for grabbing the settings from an existing training/dataset to reuse its settings. I haven't gotten around to adding it to the…
I'll take a look when I can. If my assumption is correct that you did the "run all blocks in sequence" thing, then it needs the folder to be created after you mount the Drive (which is odd, since…
didn't overwrite for me, I still had to copy manually.
Noted, I'll dig around for the right flags to assume overwrite (although I thought that was what /Y did). I guess that might explain…
mmm
I'm not too sure what would cause ROCm specific errors. Are you using the default model or a finetuned one? I'd imagine a finetuned one might throw some issues, but I wouldn't know.
(Finally) added in commit f6d0b66e1009d5a8c797219761eed1033a4735ce.
It'll search for the finetune folders in the ./training/ folder, so you won't even need to copy them now.
Did you copy the files from .\ai-voice-cloning\dlas\bitsandbytes_windows\ into .\ai-voice-cloning\venv\Lib\site-packages\bitsandbytes\?
The setup-cuda.sh script should do it automagically…
Remedied in commit 1e0fec4358a11bac2cf20cef05d3e890638ee7c5. I renamed it to "Print/Save Frequency (in epochs)".
No you're right, it implies it'll do it within an epoch, rather than every epoch. desu it was originally labeled "Print Frequency" and "Save Frequency" when it was just in terms of iterations, but…