-
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
8837bc34d7
added option to specify parameters to freeze per-model in YAML (because I need to see about committing atrocities with convering an AR into an AR+NAR)
is this because the input is too long should I split it up with the line Delimiter?
No stop tokens found in one of the generated voice clips. This typically means the spoken audio is too…
712808494f
added support for optional prodigy optimizer (https://github.com/konstmish/prodigy) although it consumes a lot more VRAM per parameter
I feel rather silly.
I imagine the lifeiteng/vall-e implementation had the right idea with having an (almost) single model that handles both AR and NAR…
100ca6b7d0
added option to use SGD optimizer through the YAML, added option to pass in additional optimizer parameters through the YAML, added experimental unified AR+NAR model (does not seem fruitful in testing)
mmm... I guess I'm due for a bit of an update.
- 15% through my dataset's "epoch" (out of ~5640894 samples, I think I should have a better metric that should calculate how many iterations it will…
451726fdd5
added ability to disable activation checkpointing through the YAML (it is very VRAM intensive at double layer size)
DeepSpeed isn't available on Windows natively without going through a nightmare of hoops to get it compiled and installed. The best bet is to have an environment under WSL2 but you're still under…
5f80ee9b38
set use-deepspeed to false because it's not a dependency and installing it as a dependency under windows is a huge nightmare
