debug: 修复了多个bug

1.完成了端侧发送杂音识别
2.修复minimax返回帧解析错误bug
3.修复了返回过慢的bug

utils/vits_utils.py

@@ -9,21 +9,6 @@ from .vits import utils, commons
 from .vits.models import SynthesizerTrn
 from .vits.text import text_to_sequence
 
-def tts_model_init(model_path='./vits_model', device='cuda'):
-    hps_ms = utils.get_hparams_from_file(os.path.join(model_path, 'config.json'))
-    # hps_ms = utils.get_hparams_from_file('vits_model/config.json')
-    net_g_ms = SynthesizerTrn(
-        len(hps_ms.symbols),
-        hps_ms.data.filter_length // 2 + 1,
-        hps_ms.train.segment_size // hps_ms.data.hop_length,
-        n_speakers=hps_ms.data.n_speakers,
-        **hps_ms.model)
-    net_g_ms = net_g_ms.eval().to(device)
-    speakers = hps_ms.speakers
-    utils.load_checkpoint(os.path.join(model_path, 'G_953000.pth'), net_g_ms, None)
-    # utils.load_checkpoint('vits_model/G_953000.pth', net_g_ms, None)
-    return hps_ms, net_g_ms, speakers
-
 class TextToSpeech:
     def __init__(self, 
                  model_path="./utils/vits_model", 
@@ -36,6 +21,11 @@ class TextToSpeech:
         self.device = torch.device(device)
         self.limitation = os.getenv("SYSTEM") == "spaces"  # 在huggingface spaces中限制文本和音频长度
         self.hps_ms, self.net_g_ms, self.speakers = self._tts_model_init(model_path)
+        self._init_jieba()
+    
+    def _init_jieba(self):
+        text = self._preprocess_text("初始化", 0)
+        self._generate_audio(text, 100, 0.6, 0.668, 1.0)
         
     def _tts_model_init(self, model_path):
         hps_ms = utils.get_hparams_from_file(os.path.join(model_path, 'config.json'))