• 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
mrq pushed to master at mrq/vall-e 2023-09-10 03:26:17 +00:00
c74fe2f718 tweaks to web UI
mrq pushed to master at mrq/vall-e 2023-09-10 01:36:57 +00:00
mrq pushed to master at mrq/vall-e 2023-09-10 01:16:38 +00:00
ed135a25f5 disable AMP in inference (for now(?))
mrq pushed to master at mrq/vall-e 2023-09-10 01:03:55 +00:00
7f8bd2b936 added printing elasped inference time
mrq pushed to master at mrq/vall-e 2023-09-09 23:03:36 +00:00
4f61f5c889 added option to set the trim length for an input prompt
mrq commented on issue mrq/ai-voice-cloning#152 2023-09-09 22:37:30 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

Before I go fuck off and let the models train for however long, a status update (in no particular order of importance):

mrq pushed to master at mrq/vall-e 2023-09-09 22:03:41 +00:00
d10053d11f render README.md markdown for huggingface space
mrq pushed to master at mrq/vall-e 2023-09-09 21:50:35 +00:00
bc30026377 added advanced sampler parameters to the web UI
mrq pushed to master at mrq/vall-e 2023-09-09 21:16:16 +00:00
5ac119a6e7 added light web UI (need to port the telemetry disabling bandaids from aivc)
mrq pushed to master at mrq/vall-e 2023-09-09 21:07:18 +00:00
eef32e4923 fixes for pure cpu
mrq pushed to master at mrq/vall-e 2023-09-09 20:02:31 +00:00
0053ae690a more tweaks
mrq pushed to master at mrq/vall-e 2023-09-09 19:46:08 +00:00
f92a20cdc1 tweaks
mrq pushed to master at mrq/vall-e 2023-09-09 19:28:00 +00:00
46d6b97eda added light web UI (need to port the telemetry disabling bandaids from aivc)
mrq pushed to master at mrq/vall-e 2023-09-09 02:00:54 +00:00
10c34c5b98 added a length-based decay factor for repetition penalty
mrq pushed to master at mrq/vall-e 2023-09-09 01:42:24 +00:00
b922f35b6b added documentation on how these new sampling parameters are very iffy and you really need to know what you are doing to use them because this is audio generation and not text generation
mrq pushed to master at mrq/vall-e 2023-09-09 01:29:58 +00:00
14c78bae39 added lots of sampling options (top-k/top-p, repetition penalty, length penalty)
mrq pushed to master at mrq/vall-e 2023-09-08 20:35:32 +00:00
f69aad9c65 some day I'll get it right
mrq pushed to master at mrq/vall-e 2023-09-08 06:02:16 +00:00
b2907ae7e0 seems that my PromEmbedding/RespEmbedding doesn't actually work all that well, naively using dedicated MultiEmbeddings for AR/NAR in the monolithic model is the best way to go
mrq commented on issue mrq/ai-voice-cloning#152 2023-09-08 00:15:30 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

I think I'm pretty pilled on using a monolithic AR+NAR.

I was training a half-sized monolithic model on the side (also making use of prodigy, major…

mrq pushed to master at mrq/vall-e 2023-09-07 23:25:48 +00:00
67617d7d69 also cull frozen_params in the params optimizer receives to reduce VRAM it consumes