SWivid
|
d3badb95cf
|
fp16 inference only for cuda devices now
|
2024-10-21 03:34:28 +08:00 |
|
SWivid
|
b4f81425f3
|
disable fp16 for cpu device
|
2024-10-20 22:45:54 +08:00 |
|
SWivid
|
28b46d32d3
|
rewrite without einx & einops; clean up
|
2024-10-20 20:41:24 +08:00 |
|
SWivid
|
554f3189e1
|
Use default fp16 inference
|
2024-10-20 16:32:18 +08:00 |
|
SWivid
|
69850fa236
|
fix. address #179
|
2024-10-20 12:43:01 +08:00 |
|
Jarod Mica
|
2a521e9050
|
Merge branch 'SWivid:main' into main
|
2024-10-15 22:58:31 -07:00 |
|
SWivid
|
bc6331529a
|
split pkgs only for eval usage address #97; clean-up
|
2024-10-15 21:14:44 +08:00 |
|
Jarod Mica
|
31e5051d51
|
default to weights_only=True for safer loading
|
2024-10-15 00:37:46 -07:00 |
|
Jarod Mica
|
6fda7e5f6f
|
Update to make passing in custom paths easier for finetuning/training
|
2024-10-14 20:13:07 -07:00 |
|
SWivid
|
e938b40bee
|
add more detailed instruct. on inference. address #49 #50
|
2024-10-14 10:15:40 +08:00 |
|
SWivid
|
615d183a0d
|
add code-switch friendly synth. and a smoother silence remover
|
2024-10-14 00:29:30 +08:00 |
|
SWivid
|
83fbd34dc8
|
convert all input audio to mono
|
2024-10-13 13:39:16 +08:00 |
|
SWivid
|
9395289d7a
|
add ckpt load opt. for .safetensor
|
2024-10-13 10:55:18 +08:00 |
|
SWivid
|
39ce201c4e
|
disable mask for single infer to save mem; add custom trans for vocab to address oov
|
2024-10-10 17:05:39 +08:00 |
|
SWivid
|
074881635d
|
basic
|
2024-10-08 21:56:51 +08:00 |
|