blackteay/GPT-SoVITS
1
0
Fork 0
mirror of https://github.com/RVC-Boss/GPT-SoVITS.git synced 2025-08-23 21:19:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

清理无用代码

Browse Source
This commit is contained in:
ChasonJiang 2025-05-05 20:16:59 +08:00
parent ab7589b5b4
commit e3d5d07064
1 changed files with 2 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
GPT_SoVITS/TTS_infer_pack/TTS.py
View File

@ -1330,22 +1330,12 @@ class TTS:
continue continue
_semantic_tokens = semantic_tokens _semantic_tokens = semantic_tokens
# if is_first_chunk:
# _semantic_tokens = torch.cat([torch.ones((1,overlap_length), dtype=torch.long, device=self.configs.device)*self.configs.mute_tokens[self.configs.version], _semantic_tokens], dim=-1)
# else:
# _semantic_tokens = torch.cat([last_tokens[:, -overlap_length:], _semantic_tokens], dim=-1)
# # _semantic_tokens = torch.cat(previous_tokens+[_semantic_tokens,], dim=-1)
previous_tokens.append(semantic_tokens) previous_tokens.append(semantic_tokens)
_semantic_tokens = torch.cat(previous_tokens, dim=-1) _semantic_tokens = torch.cat(previous_tokens, dim=-1)
# last_tokens = semantic_tokens
# print(f"_semantic_tokens shape:{_semantic_tokens.shape}")
if not self.configs.use_vocoder: if not self.configs.use_vocoder:
audio_chunk = self.vits_model.decode( audio_chunk = self.vits_model.decode(
_semantic_tokens.unsqueeze(0), _semantic_tokens.unsqueeze(0),
@ -1361,12 +1351,6 @@ class TTS:
result_length = semantic_tokens.shape[-1]+overlap_length if not is_first_chunk else None result_length = semantic_tokens.shape[-1]+overlap_length if not is_first_chunk else None
) )
# if is_first_chunk:
# audio_chunk = audio_chunk[overlap_size:]
# # is_first_chunk = False
audio_chunk_ = audio_chunk audio_chunk_ = audio_chunk
if is_first_chunk and not is_final: if is_first_chunk and not is_final:
is_first_chunk = False is_first_chunk = False
@ -1379,7 +1363,7 @@ class TTS:
audio_chunk_[last_audio_chunk.shape[0]-overlap_size:-overlap_size] if not is_final \ audio_chunk_[last_audio_chunk.shape[0]-overlap_size:-overlap_size] if not is_final \
else audio_chunk_[last_audio_chunk.shape[0]-overlap_size:] else audio_chunk_[last_audio_chunk.shape[0]-overlap_size:]
) )
# audio_chunk_ = audio_chunk_[:-overlap_size] if not is_final else audio_chunk_
last_audio_chunk = audio_chunk last_audio_chunk = audio_chunk
yield self.audio_postprocess( yield self.audio_postprocess(
@ -1391,7 +1375,7 @@ class TTS:
0.0, 0.0,
super_sampling if self.configs.use_vocoder and self.configs.version == "v3" else False, super_sampling if self.configs.use_vocoder and self.configs.version == "v3" else False,
) )
print(f"first_package_delay: {time.perf_counter()-t0:.3f}") # print(f"first_package_delay: {time.perf_counter()-t0:.3f}")
yield output_sr, np.zeros(int(output_sr*fragment_interval), dtype=np.int16) yield output_sr, np.zeros(int(output_sr*fragment_interval), dtype=np.int16)