Update src/mindomni.py

src/mindomni.py

@@ -153,9 +153,9 @@ class MindOmni:
             images=image_inputs,
             videos=video_inputs,
             padding=True,
-            return_tensors="pt",
+            return_tensors='pt',
         )
-        inputs = inputs.to("npu")
+        inputs = inputs.to(self.device)
 
         if use_cot:
             # Inference: Generation of the output