Compare commits

...

5 Commits

Author SHA1 Message Date
Masoud Azizi
b5e70761e8
Merge e8e794daa41ec71344611f6531c1b54b335c8cbb into c767f0b83b998e996a4d230d86da575a03f54a3f 2026-01-21 15:19:23 +00:00
ChasonJiang
c767f0b83b
修复bug (#2704)
* 修复bug

* fallbak and bug fix
2025-12-30 16:00:21 +08:00
ChasonJiang
9080a967d5
修复采样错误 (#2703) 2025-12-30 15:21:03 +08:00
Masoud
e8e794daa4
Update README.md
Pashto from Afghanistan is an important language that no one seems seen it. I cant found any data to train and I think a 1min trainer tool will be  very usefull for me
2024-03-20 08:09:52 +03:30
Masoud
fa6c5aadf3
Update README.md 2024-03-20 07:59:39 +03:30
2 changed files with 6 additions and 2 deletions

View File

@ -707,12 +707,11 @@ class Text2SemanticDecoder(nn.Module):
if idx == 0:
attn_mask = F.pad(attn_mask[:, :, -1].unsqueeze(-2), (0, 1), value=False)
logits = logits[:, :-1]
else:
attn_mask = F.pad(attn_mask, (0, 1), value=False)
if idx < 11: ###至少预测出10个token不然不给停止0.4s
logits = logits[:, :-1]
logits = logits[:, :-1]
samples = sample(
logits, y, top_k=top_k, top_p=top_p, repetition_penalty=repetition_penalty, temperature=temperature

View File

@ -373,6 +373,11 @@ Use v2Pro from v1/v2/v3/v4 environment:
- [x] Localization in Japanese and English.
- [x] User guide.
- [x] Japanese and English dataset fine tune training.
- [ ] Appending other languages
- [ ] 'fa': Persian
- [ ] 'pt': Pashto
- [ ] 'tr': Turkish
- [ ] 'ar': Arebic
- [ ] **Features:**
- [x] Zero-shot voice conversion (5s) / few-shot voice conversion (1min).