Update app.py

app.py

@@ -55,7 +55,8 @@ model = Qwen2VLForConditionalGeneration.from_pretrained(
     destination_folder,
     torch_dtype=torch.bfloat16,
     # device_map="cpu",
-    _attn_implementation = "flash_attention_2",
+    attn_implementation="flash_attention_2",    
+    # _attn_implementation = "flash_attention_2",
 )
 
 # Load the processor