??

demo/infer.py

@@ -34,7 +34,7 @@ class LiveCCDemoInfer:
     def __init__(self, model_path: str = None, device_id: int = 0):
         self.model = Qwen2VLForConditionalGeneration.from_pretrained(
             model_path, torch_dtype="auto", 
-            # attn_implementation='sdpa'
+            attn_implementation='eager'
         )
         self.model.to(f'cuda:{device_id}')
         self.processor = AutoProcessor.from_pretrained(model_path, use_fast=False)