-
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
If we start off with a single long dialogue wav, where would we ideally want it cut?
Segment by every sentence, if:
- it's under 200 characters, as defined in the training YAML
- when…
8ed09f9b87
Merge pull request 'Catch OOM and run whisper on cpu automatically.' (#117) from zim33/ai-voice-cloning:vram into master
Ah, I don't know why I didn't think to check if there was a way to set it to CPU only for openai/whisper.
I suppose for Windows users with low GPU VRAM but high enough system VRAM, this will do…
mm, I suppose that could be one way to automatically check if segments aren't trimmed too much. I could then dump the failures into another text file to narrow down what's needed for manual…
And I spoke too soon, I got it when starting training on a fresh install and setup:
I wonder if it's a recent gradio thing. My…
It could be. I had to disable a max row limit since it was at a very conservative 5000 rows, so I guess we're both hitting it, especially with the very-much increased resolution.
Strange…
That looks more like a gradio-related issue, which there's only so much I can do about it.
Which ironically, I just checked a training on paperspace and my graphs stopped updating, but the main…
I assume I can just load the file into an audio program, and find the exact timecode in seconds, and edit the whisper.json. Then re-slice.
Yeah, that'd be the path of least resistance when…
Not possible. The GPU count doesn't get passed on Windows from the UI => train.bat => ./src/train.py. The launcher is default to none, so it won't even bother using a job launcher.
The…
Seems to have broken multi-GPU training on Windows
To be technical, there never was. I'll never be able to validate it myself for Windows, as my GPUs are two 6800XTs and a 2060.
However, I…
I thought it also happened on the original whisper?
It's the same on whisper for me
Based on going through most of my voice samples, normal whisper's timestamps do have minor accuracy…