Update UI and fix some bugs

app.py

@@ -148,20 +148,26 @@ with gr.Blocks(theme='gradio/soft') as demo:
     "'Describe this image', 'What can you see in this image?', or "
     "'Explain what's happening in this image'."
     )
-    query = gr.Textbox(
-        lines=2, 
-        label='Chat Input', 
-        placeholder='Type your question or statement here, or upload an image and ask about it...', 
-        hint='E.g., "Describe this image" or "What is the capital of France?"'
-    )
-    task_history = gr.State([])
 
+    with gr.Row():  # Erste Zeile für Eingabefelder
+        with gr.Column(width=6):
+            query = gr.Textbox(
+                lines=2, 
+                label='Chat Input', 
+                placeholder='Type your question or statement here, or upload an image and ask about it...', 
+                hint='E.g., "Describe this image" or "What is the capital of France?"'
+            )
+        task_history = gr.State([])
+        with gr.Column(width=6):
+            upload_btn = gr.File(label="🖼️ Upload", file_types=["image"], elem_classes="control-width")
 
     with gr.Row():
-        upload_btn = gr.File(label="🖼️ Upload", file_types=["image"], elem_classes="control-width")
-        submit_btn = gr.Button("🚀 Submit", elem_classes="control-width", variant="primary")
-        regen_btn = gr.Button("🔄 Regenerate", elem_classes="control-width")
-        clear_btn = gr.Button("🧹 Clear History", elem_classes="control-width", variant="secondary")
+        with gr.Column(width=6):
+            submit_btn = gr.Button("🚀 Submit", elem_classes="control-width", variant="primary")
+        with gr.Column(width=3):
+            regen_btn = gr.Button("🔄 Regenerate", elem_classes="control-width")
+        with gr.Column(width=3):
+            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(