ChasonJiang
|
6dd2f72090
|
更改gpt并行推理时的mask策略为padding left (#2144)
* 更改gpt并行推理时的mask策略为padding left,使batch_infer更接近于naive_infer
减少冗余操作并使用torch_sdpa,以提升推理速度
* rollback tts_infer.yaml
|
2025-03-04 16:45:37 +08:00 |
|
wzy3650
|
087fd24579
|
fix inference issue (#2061)
Co-authored-by: wangzeyuan <wangzeyuan@agora.io>
|
2025-02-17 10:38:04 +08:00 |
|
wzy3650
|
282ae1d9b2
|
optimize attention calc logic (#2010)
Co-authored-by: wangzeyuan <wangzeyuan@agora.io>
|
2025-02-12 11:57:17 +08:00 |
|
RVC-Boss
|
40cd22e69d
|
Update t2s_model.py
|
2024-09-10 19:13:42 +08:00 |
|
ChasonJiang
|
52c50c6c81
|
All in one! 合并main分支和fast_inference_分支 (#1490)
* 合并main分支和fast_inference_分支
* 修复一些bug
|
2024-08-20 22:19:04 +08:00 |
|
RVC-Boss
|
3d9c37baca
|
Add files via upload
|
2024-07-08 20:52:19 +08:00 |
|
XXXXRT666
|
923af471d2
|
Update t2s_model.py (#1283)
y is None at no reference mode, and Nonetype object has no attribute shape
|
2024-07-08 20:51:15 +08:00 |
|
RVC-Boss
|
b0786f2998
|
double inference speed
double inference speed
|
2024-07-07 21:34:28 +08:00 |
|
root
|
4496426896
|
修改代码引用,淡定
|
2024-02-28 17:31:19 +08:00 |
|
Watchtower-Liu
|
013db82d7d
|
Merge branch 'main' into inference_optin
|
2024-02-16 17:03:11 +08:00 |
|
Watchtower-Liu
|
1803729360
|
update_infer
|
2024-02-16 16:53:57 +08:00 |
|
RVC-Boss
|
3bf71d3790
|
Update t2s_model.py
|
2024-02-12 16:47:43 +08:00 |
|
liufenghua
|
070ac9b2b2
|
add DPO training
|
2024-02-11 15:06:01 +08:00 |
|
RVC-Boss
|
872134c846
|
Update t2s_model.py
|
2024-01-22 23:26:18 +08:00 |
|
Blaise
|
0d3d47f3c3
|
more code refactor
|
2024-01-16 17:14:18 +01:00 |
|
RVC-Boss
|
41ca6028d6
|
Add files via upload
|
2024-01-16 17:38:48 +08:00 |
|