update

app.py

@@ -165,14 +165,10 @@ with gr.Blocks(theme='gradio/soft') as demo:
     task_history = gr.State([])
 
     with gr.Row():
-        with gr.Column(width=4):
             upload_btn = gr.UploadButton("📁 Upload", file_types=["image"], elem_classes="control-width")
-        with gr.Column(width=2):
-            submit_btn = gr.Button("🚀 Submit", elem_classes="control-width")
-        with gr.Column(width=2):
+            submit_btn = gr.Button("🚀 Submit", elem_classes="control-width", variant="primary")
             regen_btn = gr.Button("🤔️ Regenerate", elem_classes="control-width")
-        with gr.Column(width=2):
-            clear_btn = gr.Button("🧹 Clear History", elem_classes="control-width")
+            clear_btn = gr.Button("🧹 Clear History", elem_classes="control-width", variant="secondary")
     
     gr.Markdown("### Key Features:\n- **Strong Performance**: Surpasses existing LVLMs on multiple English benchmarks including Zero-shot Captioning and VQA.\n- **Multi-lingual Support**: Supports English, Chinese, and multi-lingual conversation.\n- **High Resolution**: Utilizes 448*448 resolution for fine-grained recognition and understanding.")
     submit_btn.click(handle_text_input, [chatbot, task_history, query], [chatbot, task_history]).then(
@@ -183,7 +179,8 @@ with gr.Blocks(theme='gradio/soft') as demo:
     clear_btn.click(clear_history, [task_history], [chatbot], show_progress=True)
     regen_btn.click(handle_regeneration, [chatbot, task_history], [chatbot], show_progress=True)
     upload_btn.upload(handle_file_upload, [chatbot, task_history, upload_btn], [chatbot, task_history], show_progress=True)
-    audio.on_change(transcribe_audio, inputs=[audio], outputs=[query])
+    audio.change(transcribe_audio, inputs=[audio], outputs=[query])
+
 
 
 demo.launch()