Upload app.py

app.py

@@ -128,12 +128,17 @@ def run_model(prompt: str, image: Optional[Image.Image], system_prompt: Optional
         if torch.cuda.is_available():
             model_kwargs["device_map"] = "auto"
         model = AutoModelForImageTextToText.from_pretrained(MODEL_ID, **model_kwargs)
-        # Create a pipeline for convenience
+        # Apply our generation configuration directly to the model.  The
+        # transformers pipeline does not accept a ``generation_config``
+        # keyword argument, so we assign it here.  See the Spaces
+        # documentation for more details on setting generation options.
+        model.generation_config = gen_cfg
+        # Create a pipeline for convenience.  The pipeline will use the
+        # model's generation_config when invoked.
         vlm = pipeline(
             task="image-text-to-text",
             model=model,
             processor=processor,
-            generation_config=gen_cfg,
         )
         # Store for reuse
         run_model.model = model