-
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
You might need to use either an older version of torch-directml or transformers. I don't have my previous venv of DirectML still around unfortunately, but you can start with doing (sourced from…
Added in commit e2cd07d5602742ec4068195532afa2ec4d56bebe.
e227ab8e08
updated whisperX integration for use with the latest version (v3) (NOTE: you WILL need to also update whisperx if you pull this commit)
This shouldn't be an issue, since both datasets provide speakers' Librivox unique speaker ID.
Oh duh, why didn't I think of that. I can probably make do with merging speakers but not book IDs…
There's a relatively new TTS called Balacoon, aimed at low end devices claiming 50x inference speed improvement on CPUs
Neat.
talks about some of the optimizations they made to do…
Consider replacing LibriTTS with this if you haven't
Shit, I forgot to do this. Gomen. I'm a bit tied up for the rest of the week, so...
You might get lucky with just copying an existing do_tts.py (such as from 152334H's fork, I think it…
Since it's been months since I actually touched anything with the TorToiSe training code, you might need to grab 152334H's [frozen requirements.txt](https://github.com/152334H/DL-Art-School/blob/ma…
Could have sworn I sent a post here, but I suppose I didn't.
Training is slowly improving over the weekend with a maintained LR of 1.0e-4 on the full-size model; but I don't know if I should…
OK so I have a 2000 hour audiobook dataset compiled. Didn't take that long to gather but uploading it took forever. It's untranscribed still as well.
Sweet. Training seemed to have slow down…
From what I remember, the "iterations" in the web UI determines how many steps to run the outputted codes from the AR through the diffusion sampler (desu the names I went with are a bit confusing).…
But I downloaded the current whisperx repo and there is no 'transcribe_with_vad' phrase anywhere in it. According to grep
Oh right, I forgot. I think it was v3 is when he broke everything…
if you produce a TTS clip let say on Tortoise and then feed that into RVC the quality is drastically improved
It has crossed my mind to just have TorToiSe (or the other VALL-E homebrewed…
Under the training configuration generator sub-tab:
Resume State Path: the last training state saved to resume from. The general path structure is what the placeholder value is. This will…
https://github.com/facebookresearch/fairseq/blob/main/examples/mms
The ASR might be promising at the very least, but I'm not too sure if it'd be comparable to Whisper in terms of timestamped…
I honestly can't quite recall the intricacies of TorToiSe finetuning (as most of that knowledge emptied out for VALL-E). But I sort of remember with my Japanese dataset (which I checked finally…