forked from mrq/tortoise-tts
e650800447
Noticed that the autoregressive batch size was being set off of VRAM size. Adjusted to scale for the VRAM capacity of 90 series GPUs. In this case, 16 -> 32 batches. Using the standard pre-set with ChungusVGAN, I went from 16 steps to 8. Over an average of 3 runs, I achieved an average of 294 seconds with 16 batches, to 234 seconds with 32. Can't complain at a 1.2x speed increase with functionally 2 lines of code. Can't complain. I restarted tortoise each run, and executing ```torch.cuda.empty_cache()``` just before loading the autoregressive model to clean the memory cache each time. |
||
---|---|---|
config | ||
convert | ||
scripts | ||
tortoise | ||
.gitignore | ||
CITATION.cff | ||
LICENSE | ||
MANIFEST.in | ||
README.md | ||
requirements.txt | ||
requirements_legacy.txt | ||
setup.py |
README.md
(QoL improvements for) TorToiSe
This repo is for my modifications to neonbjb/tortoise-tts. If you need the original README, refer to the original repo.
> w-where'd everything go?
Please migrate to mrq/ai-voice-cloning, as that repo is the more cohesive package for voice cloning.