Update cosyvoice/cli/cosyvoice.py

cosyvoice/cli/cosyvoice.py

@@ -18,6 +18,8 @@ from modelscope import snapshot_download
 from cosyvoice.cli.frontend import CosyVoiceFrontEnd
 from cosyvoice.cli.model import CosyVoiceModel
 
+import spaces
+
 class CosyVoice:
 
     def __init__(self, model_dir):
@@ -52,6 +54,7 @@ class CosyVoice:
             tts_speeches.append(model_output['tts_speech'])
         return {'tts_speech': torch.concat(tts_speeches, dim=1)}
 
+    @spaces.GPU
     def inference_zero_shot(self, tts_text, prompt_text, prompt_speech_16k):
         prompt_text = self.frontend.text_normalize(prompt_text, split=False)
         tts_speeches = []
@@ -61,6 +64,7 @@ class CosyVoice:
             tts_speeches.append(model_output['tts_speech'])
         return {'tts_speech': torch.concat(tts_speeches, dim=1)}
 
+    @spaces.GPU
     def inference_cross_lingual(self, tts_text, prompt_speech_16k):
         if self.frontend.instruct is True:
             raise ValueError('{} do not support cross_lingual inference'.format(self.model_dir))