Update app.py

app.py

@@ -29,7 +29,7 @@ model = AutoModelForVision2Seq.from_pretrained(
     MODEL_ID,
     token=hf_token,
     torch_dtype=torch.bfloat16,
-    _attn_implementation="flash_attention_2" if DEVICE == "cuda" else "eager",
+    # _attn_implementation="flash_attention_2" if DEVICE == "cuda" else "eager",
 ).to("cuda")  # Ensure model loads on CUDA for Spaces
 
 @spaces.GPU  # Add spaces.GPU decorator for GPU acceleration