Update app.py

app.py

@@ -29,9 +29,9 @@ def generate_text(audio):
     print(sr, data.dtype)
     input_dict = asr_processor(data, sampling_rate=sr, return_tensors="pt", padding=True)
     
-    # Ensure that the input tensor is moved to the correct device (GPU or CPU)
-    logits = asr_model(input_dict.input_values.to(device)).logits
-    pred_ids = torch.argmax(logits, dim=-1)[0]
+    with torch.no_grad():
+        outputs = model(**inputs).logits
+    pred_ids = torch.argmax(outputs, dim=-1)[0]
 
     # Decode the prediction
     return asr_processor.decode(pred_ids)