mirror of
https://github.com/FunAudioLLM/CosyVoice.git
synced 2026-02-04 09:29:25 +08:00
fix lint
This commit is contained in:
@@ -156,9 +156,9 @@ class CosyVoice2(CosyVoice):
|
||||
'{}/spk2info.pt'.format(model_dir),
|
||||
configs['allowed_special'])
|
||||
self.sample_rate = configs['sample_rate']
|
||||
if torch.cuda.is_available() is False and (load_jit is True or load_trt is True or fp16 is True):
|
||||
load_jit, load_trt, fp16 = False, False, False
|
||||
logging.warning('no cuda device, set load_jit/load_trt/fp16 to False')
|
||||
if torch.cuda.is_available() is False and (load_jit is True or load_trt is True or load_vllm is True or fp16 is True):
|
||||
load_jit, load_trt, load_vllm, fp16 = False, False, False, False
|
||||
logging.warning('no cuda device, set load_jit/load_trt/load_vllm/fp16 to False')
|
||||
self.model = CosyVoice2Model(configs['llm'], configs['flow'], configs['hift'], fp16)
|
||||
self.model.load('{}/llm.pt'.format(model_dir),
|
||||
'{}/flow.pt'.format(model_dir),
|
||||
@@ -174,11 +174,7 @@ class CosyVoice2(CosyVoice):
|
||||
self.fp16)
|
||||
del configs
|
||||
|
||||
def inference_instruct(self, *args, **kwargs):
|
||||
raise NotImplementedError('inference_instruct is not implemented for CosyVoice2!')
|
||||
|
||||
def inference_instruct2(self, tts_text, instruct_text, prompt_wav, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
|
||||
assert isinstance(self.model, CosyVoice2Model) or isinstance(self.model, CosyVoice3Model), 'inference_instruct2 is only implemented for CosyVoice2 and CosyVoice3!'
|
||||
for i in tqdm(self.frontend.text_normalize(tts_text, split=True, text_frontend=text_frontend)):
|
||||
model_input = self.frontend.frontend_instruct2(i, instruct_text, prompt_wav, self.sample_rate, zero_shot_spk_id)
|
||||
start_time = time.time()
|
||||
|
||||
@@ -436,4 +436,4 @@ class CosyVoice3Model(CosyVoice2Model):
|
||||
tts_speech, _ = self.hift.inference(speech_feat=tts_mel, finalize=finalize)
|
||||
tts_speech = tts_speech[:, self.hift_cache_dict[uuid]['speech_offset']:]
|
||||
self.hift_cache_dict[uuid]['speech_offset'] += tts_speech.shape[1]
|
||||
return tts_speech
|
||||
return tts_speech
|
||||
|
||||
Reference in New Issue
Block a user