Update app.py

app.py

@@ -15,8 +15,8 @@ def transcribe(audio):
     if not isinstance(audio_data, np.ndarray):
         audio_data = np.array(audio_data)
     
-    # Convert to a tensor
-    audio_tensor = torch.tensor(audio_data)
+    # Convert to a tensor and ensure it's a floating-point type
+    audio_tensor = torch.tensor(audio_data, dtype=torch.float32)
     
     # Convert to mono if the audio is stereo
     if audio_tensor.ndim > 1:
@@ -39,6 +39,8 @@ def transcribe(audio):
     
     prediction = pipe(audio_tensor)["text"]
     return prediction
+
+    
 gradio_app = gr.Interface(
     fn=transcribe,
     inputs=gr.Audio(label="Input"),