test app.py

cosyvoice/cli/model.py

@@ -260,7 +260,7 @@ class CosyVoice2Model(CosyVoiceModel):
         # speech fade in out
         self.speech_window = np.hamming(2 * self.source_cache_len)
         # rtf and decoding related
-        self.llm_context = torch.cuda.stream(torch.cuda.Stream(self.device)) if torch.cuda.is_available() else nullcontext()
+        self.llm_context = nullcontext()
         self.lock = threading.Lock()
         # dict used to store session related variable
         self.tts_speech_token_dict = {}