Update app.py

app.py

@@ -130,7 +130,7 @@ model_v = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_V,
     trust_remote_code=True,
     torch_dtype=torch.float16,
-    _attn_implementation="flash_attention_2"
+    #_attn_implementation="flash_attention_2"
 ).to(device).eval()
 
 # Load Dots.OCR