cover

inference/slicer.py

@@ -134,7 +134,7 @@ def chunks2audio(input_audio, chunks):
     sr, audio = input_audio
     if len(audio.shape) == 2 and audio.shape[1] >= 2:
         audio = torch.mean(audio, dim=0).unsqueeze(0)
-    audio = audio.cpu().numpy()[0]
+    audio = audio.numpy()[0]
     result = []
     for k, v in chunks.items():
         tag = v["split_time"].split(",")