ecker
18 Followers · 0 Following

  • 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
Repositories
10
 Projects Packages Public Activity Starred Repositories
ecker pushed to master at ecker/tortoise-tts 2023-05-04 23:00:07 +00:00
086aad5b49 quick hotfix to remove offending codesmell (will actually clean it when I finish eating)
ecker pushed to main at ecker/tortoise-tts 2023-05-04 23:00:05 +00:00
086aad5b49 quick hotfix to remove offending codesmell (will actually clean it when I finish eating)
ecker commented on issue ecker/ai-voice-cloning#229 2023-05-04 22:56:04 +00:00
TypeError: Progress.tqdm() got an unexpected keyword argument 'track_tqdm'

I'll look into it. I'm guessing gradio broke the API. I suppose the better thing to do is to undo my nasty kludge rather than bandaid it with freezing the gradio version used.

ecker pushed to master at ecker/vall-e 2023-05-04 05:17:34 +00:00
a3045426f5 actually fixed loading model weights (I guess it works, it expects the weights to be saved at './dir/with/your/yaml/ckpt/(n)ar/fp32.pth'
ecker commented on issue ecker/ai-voice-cloning#222 2023-05-04 02:19:18 +00:00
Adding faster-whisper backend

So:

ecker pushed to main at ecker/tortoise-tts 2023-05-04 01:54:59 +00:00
04b7049811 freeze numpy to 1.23.5 because latest version will moan about deprecating complex
ecker pushed to master at ecker/tortoise-tts 2023-05-04 01:54:56 +00:00
04b7049811 freeze numpy to 1.23.5 because latest version will moan about deprecating complex
ecker commented on issue ecker/ai-voice-cloning#223 2023-05-03 22:35:26 +00:00
Trying to Transcribe but getting error: 'Missing dataset: ./training/voice1//whisper.json'

I might have incidentally fixed this in commit 853c7fdccf13277e500307aa241e149d7e99ac7d, if you were using "Transcribe All" instead (which transcribes ALL voices, rather than the selected…

ecker pushed to master at ecker/vall-e 2023-05-03 21:34:28 +00:00
43d004923b added reporting of training precision, (need to test) load raw model weights, (need to test) dockerfile for whenever I throw this bad boy onto a runpod machine again
ecker pushed to master at ecker/ai-voice-cloning 2023-05-03 21:31:42 +00:00
853c7fdccf forgot to uncomment the block to transcribe and slice when using transcribe all because I was piece-processing a huge batch of LibriTTS and somehow that leaked over to the repo
ecker commented on issue ecker/ai-voice-cloning#228 2023-05-03 20:15:21 +00:00
Bug/Issue: Incorrect save steps passed to train.yaml

125 lines batch size 64

125 lines / 64 lines per batch will yield 2 iterations per epoch.

DLAS takes its inputs as iteration counts rather than epochs.

ecker commented on issue ecker/ai-voice-cloning#152 2023-05-03 02:39:54 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

Progress report: since my metrics seemed to have flatlined after running through the LR schedule, I went ahead and:

  • swapped to the bigger portion of the dataset that got culled (phoneme lengths…
ecker pushed to master at ecker/vall-e 2023-05-03 01:47:52 +00:00
6db57d7697 added config option to directly set the step counter (to avoid needing to ditch optimizer states
ecker commented on issue ecker/ai-voice-cloning#219 2023-04-29 14:07:43 +00:00
mrq adding significant American accent to same voice samples from tortoise-fast-tts

Number of Diffusion Steps should map to Iterations (because it's the amount of iterations taken of diffusion to create the waveform).

autoregressive samples should map to Samples…

ecker commented on issue ecker/ai-voice-cloning#220 2023-04-29 03:03:06 +00:00
Import/Analyse Processing Infinitely & fills up appdata/temp

I've currently resorted to just making a folder with the samples manually.

You're suppose to add voices [that way](https://git.ecker.tech/mrq/ai-voice-cloning/wiki/Collecting-Samples#preparing…

ecker pushed to master at ecker/ai-voice-cloning 2023-04-29 00:50:25 +00:00
fd306d850d updated setup-directml.bat to not hard require torch version because it's updated to torch2 now
ecker pushed to master at ecker/ai-voice-cloning 2023-04-29 00:47:17 +00:00
a755526fec updated setup-directml.bat to not hard require torch version because it's updated to torch2 now
ecker pushed to master at ecker/tortoise-tts 2023-04-29 00:46:46 +00:00
b6a213bbbd removed some CPU fallback wrappers because directml seems to work now without them
ecker pushed to main at ecker/tortoise-tts 2023-04-29 00:46:43 +00:00
b6a213bbbd removed some CPU fallback wrappers because directml seems to work now without them
ecker commented on issue ecker/ai-voice-cloning#219 2023-04-29 00:40:00 +00:00
mrq adding significant American accent to same voice samples from tortoise-fast-tts

I’m actually not sure if that one produces a pt file as yours does to be honest I’ve tried to figure it out but I cant find any corresponding files.

From what I remember, the 152334H…