Update app.py

app.py

@@ -28,7 +28,6 @@ except subprocess.CalledProcessError as e:
     print(f"Error installing flash-attn: {e}")
     print("Continuing without flash-attn.")
 
-import flash_attn_2_cuda as flash_attn_gpu
 
 colors.steel_blue = colors.Color(
     name="steel_blue",
@@ -159,6 +158,8 @@ model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
+import flash_attn_2_cuda as flash_attn_gpu
+
 # Load Dots.OCR from the local, patched directory
 MODEL_PATH_D = model_path_d_local
 processor_d = AutoProcessor.from_pretrained(MODEL_PATH_D, trust_remote_code=True)