speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

cut paragraph for streaming s2s

Browse Source
This commit is contained in:
志浩 2024-09-12 19:51:43 +08:00
parent 77b75ae3b2
commit 4b65ccee2a
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
funasr/models/llm_asr/model.py
View File

@ -3031,7 +3031,7 @@ class LLMASRXvecSlotTTS(nn.Module):
idx = -1
for p in pounc:
str_idx = normed_preds.find(p)
if str_idx > 0:
if str_idx > -1:
preds = self.split_characters_and_words(normed_preds[:str_idx])
idx = len(preds)
preds.append(normed_preds[str_idx])
@ -3096,6 +3096,8 @@ class LLMASRXvecSlotTTS(nn.Module):
last_t_size = 0
prompt_token, prompt_audio = [None, None], [None, None]
wav = torch.cat([wav, torch.zeros([1, 4410]).to(wav)], dim=1)
else:
text = text + normed_preds
return ((cur_token, feat, wav), (text, last_t_size, prompt_token, prompt_audio, chunk_idx))