modified

app.py

@@ -215,18 +215,6 @@ def toggle_media_input(choice: str) -> Tuple:
     else:  # Video
         return gr.update(visible=False, value=None), gr.update(visible=True, value=None), gr.update(visible=False), gr.update(visible=True)
 
-# # --- MODIFIED: New function to handle chat state and input clearing ---
-# def process_and_clear(chatbot: List, image_input: PIL.Image.Image, video_input: str, prompt: str, do_sample: bool, max_new_tokens: int, enable_thinking: bool):
-#     """
-#     This function now takes the chatbot state as input to maintain conversation history
-#     and clears the prompt box after submission.
-#     """
-#     # Create a generator by calling the main run_inference function
-#     generator = run_inference(chatbot, image_input, video_input, prompt, do_sample, max_new_tokens, enable_thinking)
-#     # Yield from the generator
-#     for chatbot_state, _ in generator:
-#         yield chatbot_state, "" # Clear prompt after first yield
-
 
 # --- Build Gradio Application ---
 # @spaces.GPU
@@ -282,7 +270,7 @@ def build_demo(model_path: str):
 
                 with gr.Accordion("Generation Settings", open=True):
                     do_sample = gr.Checkbox(label="Enable Sampling (Do Sample)", value=True)
-                    max_new_tokens = gr.Slider(minimum=32, maximum=4096, value=1024, step=32, label="Max New Tokens")
+                    max_new_tokens = gr.Slider(minimum=32, maximum=4096, value=2048, step=32, label="Max New Tokens")
                     enable_thinking = gr.Checkbox(label="Enable Deep Thinking", value=False)