cuda

src/predict.py

@@ -70,7 +70,7 @@ class Predictor:
         self.textclenaer = None
         self.mean = 0
         self.std = 0
-        self.device = 'cuda:0'
+        self.device = 'cuda'
 
         self.ckpt_base = 'checkpoints/base_speakers/EN'
         self.ckpt_converter = 'checkpoints/converter'

src/se_extractor.py

@@ -12,7 +12,7 @@ model = None
 model_size = 'medium'
 def split_audio_whisper(audio_path, target_dir='processed',needs_offset=True):
     print("in whisper split")
-    model = WhisperModel('medium', device="cuda:0", compute_type="float16")
+    model = WhisperModel('medium', device="cuda", compute_type="float16")
     audio = AudioSegment.from_file(audio_path)
     max_len = len(audio)
 
@@ -162,7 +162,7 @@ def load_model():
 
 def extract_segments_to_cut_audio(max_duration,audio_path,target_dir='processed'):
     
-    model = WhisperModel('medium', device="cuda:0", compute_type="float16")
+    model = WhisperModel('medium', device="cuda", compute_type="float16")
     audio = AudioSegment.from_file(audio_path)
     max_len = len(audio)