Update app.py

app.py

@@ -16,7 +16,7 @@ def generate(model_name, image, text):
     print("Generating!")
     # Wrap the model call in a try-except block to capture and debug CUDA errors
     try:
-        result = model_pipe(images=[image], texts=[text])  # Perform the model inference
+        result = model_pipe(images=[image], texts=[text]).cpu()  # Perform the model inference
     except RuntimeError as e:
         print(f"RuntimeError during model inference: {e}")
         raise e