-
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
While that's fair, it's a bad practice desu to invoke programs/scripts that are in the CWD solely by name, like start or run (at least that's the impression I've developed over my years).
Ju…
7f98727ad5
added option to specify autoregressive model at tts generation time (for a spicy feature later)
7f98727ad5
added option to specify autoregressive model at tts generation time (for a spicy feature later)
cf9ca97c9f
added option to specify autoregressive model at tts generation time (for a spicy feature later)
cf9ca97c9f
added option to specify autoregressive model at tts generation time (for a spicy feature later)
I'll second the motion on some kind of batching system. The nature of audio work is that you get a long list of sentences to be spoken by a character and you just want the machine to munch on…
119ac50c58
forgot to re-append the existing transcription when skipping existing (have to go back again and do the first 10% of my giant dataset
Remedied in 11a1f6a00ea3a2408d7592537a748433ad3b5496. I forgot to commit it at 1AM last night.
WhisperX will break the dependencies needed for DLAS, so installing DLAS's dependencies will fix it.
It's a side-effect from installing whisperx. DLAS's dependencies need to be installed after to remedy it.
(Should be) fixed in 11a1f6a00ea3a2408d7592537a748433ad3b5496. Should be, because it's…
11a1f6a00e
forgot to reorder the dependency install because whisperx needs to be installed before DLAS
12c51b6057
Im not too sure if manually invoking gc actually closes all the open files from whisperx (or ROCm), but it seems to have gone away longside setting 'ulimit -Sn' to half the output of 'ulimit -Hn'
999878d9c6
and it turned out I wasn't even using the aligned segments, kmsing now that I have to *redo* my dataset again
14779a5020
Added option to skip transcribing if it exists in the output text file, because apparently whisperx will throw a "max files opened" error when using ROCm because it does not close some file descriptors if you're batch-transcribing or something, so poor little me, who's retranscribing his japanese dataset for the 305823042th time woke up to it partially done i am so mad I have to wait another few hours for it to continue when I was hoping to wake up to it done
Use the larger models, you'll get more accuracy at the cost of throughput speed and more VRAM consumption, or use the whisperx backend that got added not too long ago that uses wav2vec2 alignment…
Don't think it matters all that much, as I've tested before both signed-16 and float-32, and neither seem to have a perceptable difference, as it gets coerced into whatever's necessary after torch…