Update demo/infer.py

demo/infer.py

@@ -35,7 +35,7 @@ class LiveCCDemoInfer:
     def __init__(self, model_path: str = None, device: str = 'cpu'):
         self.model = Qwen2VLForConditionalGeneration.from_pretrained(
             model_path, torch_dtype="auto", 
-            # device_map=f'cuda:{device_id}', 
+            device_map=device, 
             # attn_implementation='flash_attention_2'
         )
         self.model.to(device)