-
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
63bcadcbbe
actually for real fixed incrementing filenames because i had a regex that actually only worked if candidates or lines>1, cuda now takes priority over dml if you're a nut with both of them installed because you can just specify an override anyways
Should now actually be fixed in commit 7a4460ddf087f643b8df60b4317cf9b1cf8dd581.
The new behavior will use the last indexed file + 1.
Added argument device-override/setting Device Override in commit 7a4460ddf087f643b8df60b4317cf9b1cf8dd581. Pass a string returned from running list_devices.py, for example, pass cuda:1…
7a4460ddf0
added setting "device-override", less naively decide the number to use for results, some other thing
Gross, I don't want to go back to eye sore timestamps for "uniqueness". I'll admit the current way it does it is still rather naive, but I didn't seem to have any problems are extensively testing…
However there seems to be a new problem, and I'm not sure what's going on here with connections?
Those have usually happened for me. They're very innocuous so I haven't bothered with…
The system cannot find the path specified: './results/'
Well, it's missing the ./results/ folder, which I suppose I forgot to have created if not available outside of a generation call.…
88535f192d
voicefixed files do not overwrite, as my autism wants to hear the difference between them, incrementing file format fixed for real
f4d2d0d7f8
added option: force cpu for conditioning latents, for when you want low chunk counts but your GPU keeps OOMing because fuck fragmentation
defa460028
modified conversion scripts to not give a shit about bitrate and formats since torchaudio.load handles all of that anyways, and it all gets resampled anyways
2ee6068f98
done away with kludgy shit code, just have the user decide how many chunks to slice concat'd samples to (since it actually does improve vocie replicability)
I've also added a checkbox to outright disable in commit b4ca260de96a96fe54363dda399ad502befa0fe5, as it would OOM periodically on my 2060 (6GiB VRAM) due to how terribly fragmented it gets.
Added in commit c12ada600bbfccaac27594a6af78ca72b55b1969.
Reset to Defaultis just an over-glorified button to erase./config/generation.jsonor whatever it's called.Utilitiestab…