deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 09:29:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix bug

Browse Source
This commit is contained in:
lyuxiang.lx
2025-12-09 07:57:10 +00:00
parent b35ece675b
commit 56d9876037
8 changed files with 49 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
cosyvoice/cli/frontend.py
View File

@@ -193,13 +193,13 @@ class CosyVoiceFrontEnd:
model_input = self.frontend_sft(tts_text, spk_id)
# in instruct mode, we remove spk_embedding in llm due to information leakage
del model_input['llm_embedding']
instruct_text_token, instruct_text_token_len = self._extract_text_token(instruct_text + '<endofprompt>')
instruct_text_token, instruct_text_token_len = self._extract_text_token(instruct_text)
model_input['prompt_text'] = instruct_text_token
model_input['prompt_text_len'] = instruct_text_token_len
return model_input
def frontend_instruct2(self, tts_text, instruct_text, prompt_wav, resample_rate, zero_shot_spk_id):
model_input = self.frontend_zero_shot(tts_text, instruct_text + '<|endofprompt|>', prompt_wav, resample_rate, zero_shot_spk_id)
model_input = self.frontend_zero_shot(tts_text, instruct_text, prompt_wav, resample_rate, zero_shot_spk_id)
del model_input['llm_prompt_speech_token']
del model_input['llm_prompt_speech_token_len']
return model_input

6
cosyvoice/cli/model.py
View File

@@ -129,7 +129,7 @@ class CosyVoiceModel:
def token2wav(self, token, prompt_token, prompt_feat, embedding, uuid, finalize=False, speed=1.0):
with torch.cuda.amp.autocast(self.fp16):
tts_mel, self.flow_cache_dict[uuid] = self.flow.inference(token=token.to(self.device),
tts_mel, self.flow_cache_dict[uuid] = self.flow.inference(token=token.to(self.device, dtype=torch.int32),
token_len=torch.tensor([token.shape[1]], dtype=torch.int32).to(self.device),
prompt_token=prompt_token.to(self.device),
prompt_token_len=torch.tensor([prompt_token.shape[1]], dtype=torch.int32).to(self.device),
@@ -284,7 +284,7 @@ class CosyVoice2Model(CosyVoiceModel):
def token2wav(self, token, prompt_token, prompt_feat, embedding, token_offset, uuid, stream=False, finalize=False, speed=1.0):
with torch.cuda.amp.autocast(self.fp16):
tts_mel, _ = self.flow.inference(token=token.to(self.device),
tts_mel, _ = self.flow.inference(token=token.to(self.device, dtype=torch.int32),
token_len=torch.tensor([token.shape[1]], dtype=torch.int32).to(self.device),
prompt_token=prompt_token.to(self.device),
prompt_token_len=torch.tensor([prompt_token.shape[1]], dtype=torch.int32).to(self.device),
@@ -413,7 +413,7 @@ class CosyVoice3Model(CosyVoice2Model):
def token2wav(self, token, prompt_token, prompt_feat, embedding, token_offset, uuid, stream=False, finalize=False, speed=1.0):
with torch.cuda.amp.autocast(self.fp16):
tts_mel, _ = self.flow.inference(token=token.to(self.device),
tts_mel, _ = self.flow.inference(token=token.to(self.device, dtype=torch.int32),
token_len=torch.tensor([token.shape[1]], dtype=torch.int32).to(self.device),
prompt_token=prompt_token.to(self.device),
prompt_token_len=torch.tensor([prompt_token.shape[1]], dtype=torch.int32).to(self.device),