fix lint

cosyvoice/cli/cosyvoice.py

@@ -20,6 +20,7 @@ from cosyvoice.cli.frontend import CosyVoiceFrontEnd
 from cosyvoice.cli.model import CosyVoiceModel
 from cosyvoice.utils.file_utils import logging
 
+
 class CosyVoice:
 
     def __init__(self, model_dir, load_jit=True, load_onnx=True):
@@ -42,8 +43,8 @@ class CosyVoice:
                         '{}/hift.pt'.format(model_dir))
         if load_jit:
             self.model.load_jit('{}/llm.text_encoder.fp16.zip'.format(model_dir),
-                                    '{}/llm.llm.fp16.zip'.format(model_dir),
-                                    '{}/flow.encoder.fp32.zip'.format(model_dir))
+                                '{}/llm.llm.fp16.zip'.format(model_dir),
+                                '{}/flow.encoder.fp32.zip'.format(model_dir))
         if load_onnx:
             self.model.load_onnx('{}/flow.decoder.estimator.fp32.onnx'.format(model_dir))
         del configs

cosyvoice/cli/frontend.py

@@ -50,7 +50,9 @@ class CosyVoiceFrontEnd:
         option.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL
         option.intra_op_num_threads = 1
         self.campplus_session = onnxruntime.InferenceSession(campplus_model, sess_options=option, providers=["CPUExecutionProvider"])
-        self.speech_tokenizer_session = onnxruntime.InferenceSession(speech_tokenizer_model, sess_options=option, providers=["CUDAExecutionProvider"if torch.cuda.is_available() else "CPUExecutionProvider"])
+        self.speech_tokenizer_session = onnxruntime.InferenceSession(speech_tokenizer_model, sess_options=option,
+                                                                     providers=["CUDAExecutionProvider" if torch.cuda.is_available() else
+                                                                                "CPUExecutionProvider"])
         if os.path.exists(spk2info):
             self.spk2info = torch.load(spk2info, map_location=self.device)
         self.instruct = instruct
@@ -60,7 +62,8 @@ class CosyVoiceFrontEnd:
         if self.use_ttsfrd:
             self.frd = ttsfrd.TtsFrontendEngine()
             ROOT_DIR = os.path.dirname(os.path.abspath(__file__))
-            assert self.frd.initialize('{}/../../pretrained_models/CosyVoice-ttsfrd/resource'.format(ROOT_DIR)) is True, 'failed to initialize ttsfrd resource'
+            assert self.frd.initialize('{}/../../pretrained_models/CosyVoice-ttsfrd/resource'.format(ROOT_DIR)) is True, \
+                'failed to initialize ttsfrd resource'
             self.frd.set_lang_type('pinyin')
             self.frd.enable_pinyin_mix(True)
             self.frd.set_breakmodel_index(1)
@@ -76,8 +79,11 @@ class CosyVoiceFrontEnd:
 
     def _extract_speech_token(self, speech):
         feat = whisper.log_mel_spectrogram(speech, n_mels=128)
-        speech_token = self.speech_tokenizer_session.run(None, {self.speech_tokenizer_session.get_inputs()[0].name: feat.detach().cpu().numpy(),
-                                                                self.speech_tokenizer_session.get_inputs()[1].name: np.array([feat.shape[2]], dtype=np.int32)})[0].flatten().tolist()
+        speech_token = self.speech_tokenizer_session.run(None,
+                                                         {self.speech_tokenizer_session.get_inputs()[0].name:
+                                                          feat.detach().cpu().numpy(),
+                                                          self.speech_tokenizer_session.get_inputs()[1].name:
+                                                          np.array([feat.shape[2]], dtype=np.int32)})[0].flatten().tolist()
         speech_token = torch.tensor([speech_token], dtype=torch.int32).to(self.device)
         speech_token_len = torch.tensor([speech_token.shape[1]], dtype=torch.int32).to(self.device)
         return speech_token, speech_token_len
@@ -88,7 +94,8 @@ class CosyVoiceFrontEnd:
                            dither=0,
                            sample_frequency=16000)
         feat = feat - feat.mean(dim=0, keepdim=True)
-        embedding = self.campplus_session.run(None, {self.campplus_session.get_inputs()[0].name: feat.unsqueeze(dim=0).cpu().numpy()})[0].flatten().tolist()
+        embedding = self.campplus_session.run(None,
+                                              {self.campplus_session.get_inputs()[0].name: feat.unsqueeze(dim=0).cpu().numpy()})[0].flatten().tolist()
         embedding = torch.tensor([embedding]).to(self.device)
         return embedding
 
@@ -112,18 +119,16 @@ class CosyVoiceFrontEnd:
             text = text.replace(" - ", "，")
             text = remove_bracket(text)
             text = re.sub(r'[，,]+$', '。', text)
-            texts = [i for i in split_paragraph(text, partial(self.tokenizer.encode, allowed_special=self.allowed_special), "zh", token_max_n=80,
-                                                token_min_n=60, merge_len=20,
-                                                comma_split=False)]
+            texts = list(split_paragraph(text, partial(self.tokenizer.encode, allowed_special=self.allowed_special), "zh", token_max_n=80,
+                                         token_min_n=60, merge_len=20, comma_split=False))
         else:
             if self.use_ttsfrd:
                 text = self.frd.get_frd_extra_info(text, 'input')
             else:
                 text = self.en_tn_model.normalize(text)
             text = spell_out_number(text, self.inflect_parser)
-            texts = [i for i in split_paragraph(text, partial(self.tokenizer.encode, allowed_special=self.allowed_special), "en", token_max_n=80,
-                                                token_min_n=60, merge_len=20,
-                                                comma_split=False)]
+            texts = list(split_paragraph(text, partial(self.tokenizer.encode, allowed_special=self.allowed_special), "en", token_max_n=80,
+                                         token_min_n=60, merge_len=20, comma_split=False))
         if split is False:
             return text
         return texts

cosyvoice/cli/model.py

@@ -18,7 +18,7 @@ import time
 from contextlib import nullcontext
 import uuid
 from cosyvoice.utils.common import fade_in_out
-import numpy as np
+
 
 class CosyVoiceModel:
 
@@ -80,27 +80,27 @@ class CosyVoiceModel:
     def llm_job(self, text, prompt_text, llm_prompt_speech_token, llm_embedding, uuid):
         with self.llm_context:
             for i in self.llm.inference(text=text.to(self.device),
-                                                text_len=torch.tensor([text.shape[1]], dtype=torch.int32).to(self.device),
-                                                prompt_text=prompt_text.to(self.device),
-                                                prompt_text_len=torch.tensor([prompt_text.shape[1]], dtype=torch.int32).to(self.device),
-                                                prompt_speech_token=llm_prompt_speech_token.to(self.device),
-                                                prompt_speech_token_len=torch.tensor([llm_prompt_speech_token.shape[1]], dtype=torch.int32).to(self.device),
-                                                embedding=llm_embedding.to(self.device).half(),
-                                                sampling=25,
-                                                max_token_text_ratio=30,
-                                                min_token_text_ratio=3):
+                                        text_len=torch.tensor([text.shape[1]], dtype=torch.int32).to(self.device),
+                                        prompt_text=prompt_text.to(self.device),
+                                        prompt_text_len=torch.tensor([prompt_text.shape[1]], dtype=torch.int32).to(self.device),
+                                        prompt_speech_token=llm_prompt_speech_token.to(self.device),
+                                        prompt_speech_token_len=torch.tensor([llm_prompt_speech_token.shape[1]], dtype=torch.int32).to(self.device),
+                                        embedding=llm_embedding.to(self.device).half(),
+                                        sampling=25,
+                                        max_token_text_ratio=30,
+                                        min_token_text_ratio=3):
                 self.tts_speech_token_dict[uuid].append(i)
         self.llm_end_dict[uuid] = True
 
     def token2wav(self, token, prompt_token, prompt_feat, embedding, uuid, finalize=False):
         with self.flow_hift_context:
             tts_mel = self.flow.inference(token=token.to(self.device),
-                                        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),
-                                        prompt_feat=prompt_feat.to(self.device),
-                                        prompt_feat_len=torch.tensor([prompt_feat.shape[1]], dtype=torch.int32).to(self.device),
-                                        embedding=embedding.to(self.device))
+                                          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),
+                                          prompt_feat=prompt_feat.to(self.device),
+                                          prompt_feat_len=torch.tensor([prompt_feat.shape[1]], dtype=torch.int32).to(self.device),
+                                          embedding=embedding.to(self.device))
             # mel overlap fade in out
             if self.mel_overlap_dict[uuid] is not None:
                 tts_mel = fade_in_out(tts_mel, self.mel_overlap_dict[uuid], self.mel_window)
@@ -129,7 +129,8 @@ class CosyVoiceModel:
         # this_uuid is used to track variables related to this inference thread
         this_uuid = str(uuid.uuid1())
         with self.lock:
-            self.tts_speech_token_dict[this_uuid], self.llm_end_dict[this_uuid], self.mel_overlap_dict[this_uuid], self.hift_cache_dict[this_uuid] = [], False, None, None
+            self.tts_speech_token_dict[this_uuid], self.llm_end_dict[this_uuid] = [], False
+            self.mel_overlap_dict[this_uuid], self.hift_cache_dict[this_uuid] = None, None
         p = threading.Thread(target=self.llm_job, args=(text, prompt_text, llm_prompt_speech_token, llm_embedding, this_uuid))
         p.start()
         if stream is True:
@@ -140,12 +141,12 @@ class CosyVoiceModel:
                     this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid][:token_hop_len + self.token_overlap_len], dim=1)
                     with self.flow_hift_context:
                         this_tts_speech = self.token2wav(token=this_tts_speech_token,
-                                                    prompt_token=flow_prompt_speech_token,
-                                                    prompt_feat=prompt_speech_feat,
-                                                    embedding=flow_embedding,
-                                                    uuid=this_uuid,
-                                                    finalize=False)
-                    yield  {'tts_speech': this_tts_speech.cpu()}
+                                                         prompt_token=flow_prompt_speech_token,
+                                                         prompt_feat=prompt_speech_feat,
+                                                         embedding=flow_embedding,
+                                                         uuid=this_uuid,
+                                                         finalize=False)
+                    yield {'tts_speech': this_tts_speech.cpu()}
                     with self.lock:
                         self.tts_speech_token_dict[this_uuid] = self.tts_speech_token_dict[this_uuid][token_hop_len:]
                     # increase token_hop_len for better speech quality
@@ -157,11 +158,11 @@ class CosyVoiceModel:
             this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid], dim=1)
             with self.flow_hift_context:
                 this_tts_speech = self.token2wav(token=this_tts_speech_token,
-                                            prompt_token=flow_prompt_speech_token,
-                                            prompt_feat=prompt_speech_feat,
-                                            embedding=flow_embedding,
-                                            uuid=this_uuid,
-                                            finalize=True)
+                                                 prompt_token=flow_prompt_speech_token,
+                                                 prompt_feat=prompt_speech_feat,
+                                                 embedding=flow_embedding,
+                                                 uuid=this_uuid,
+                                                 finalize=True)
             yield {'tts_speech': this_tts_speech.cpu()}
         else:
             # deal with all tokens
@@ -169,11 +170,11 @@ class CosyVoiceModel:
             this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid], dim=1)
             with self.flow_hift_context:
                 this_tts_speech = self.token2wav(token=this_tts_speech_token,
-                                            prompt_token=flow_prompt_speech_token,
-                                            prompt_feat=prompt_speech_feat,
-                                            embedding=flow_embedding,
-                                            uuid=this_uuid,
-                                            finalize=True)
+                                                 prompt_token=flow_prompt_speech_token,
+                                                 prompt_feat=prompt_speech_feat,
+                                                 embedding=flow_embedding,
+                                                 uuid=this_uuid,
+                                                 finalize=True)
             yield {'tts_speech': this_tts_speech.cpu()}
         with self.lock:
             self.tts_speech_token_dict.pop(this_uuid)