update app

app.py

@@ -66,8 +66,16 @@ def get_chat_response(chatbot, task_history):
             history_filter.append((pre, a))
             pre = ""
     history, message = history_filter[:-1], history_filter[-1][0]
-    response, history = model.chat(tokenizer, message, history=history)
-    # ... (rest of the code remains the same)
+    
+    inputs = tokenizer.encode_plus(message, return_tensors='pt')
+    outputs = model.generate(inputs['input_ids'], max_length=150, num_beams=4, length_penalty=2.0, early_stopping=True)
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    
+    task_history.append((message, response))
+    chatbot.append((format_text(message), format_text(response)))
+
+    return chatbot, task_history
+
 
 def handle_text_input(history, task_history, text):
     """Handle text input from the user."""
@@ -111,40 +119,46 @@ def handle_regeneration(chatbot, task_history):
     print("After:", task_history, chatbot)
     return get_chat_response(chatbot, task_history)
 
+chatbot = []
+task_history = []
+
+def main_function(text, image):
+    global chatbot, task_history
+    if text:
+        chatbot, task_history = handle_text_input(chatbot, task_history, text)
+    if image:
+        chatbot, task_history = handle_file_upload(chatbot, task_history, image)
+    chatbot, task_history = get_chat_response(chatbot, task_history)
+    formatted_response = chatbot[-1][1]  # Get the latest response from the chatbot
+    return formatted_response
+
+def clear_history_fn():
+    global chatbot, task_history
+    chatbot.clear()
+    task_history.clear()
+    return "History cleared."
+
 # Custom CSS
 css = '''
     .gradio-container {
         max-width: 800px !important;
     }
-    /* ... (add more custom CSS if needed) */
 '''
 
-# Build and launch the UI
-with gr.Blocks(css=css) as demo:
-    gr.Markdown("# Qwen-VL-Chat Bot")
-    gr.Markdown(
-        "## Developed by Keyvan Hardani (Keyvven on [Twitter](https://twitter.com/Keyvven))\n"
-        "Special thanks to [@Artificialguybr](https://twitter.com/artificialguybr) for the inspiration from his code.\n"
-        "### Qwen-VL: A Multimodal Large Vision Language Model by Alibaba Cloud\n"
-    )
-    chatbot = gr.Chatbot(label='Qwen-VL-Chat', elem_classes="control-height", height=520)
-    query = gr.Textbox(lines=2, label='Input')
-    task_history = gr.State([])
-
-    with gr.Row():
-        upload_btn = gr.UploadButton("📁 Upload", file_types=["image"])
-        submit_btn = gr.Button("🚀 Submit")
-        regen_btn = gr.Button("🤔️ Regenerate")
-        clear_btn = gr.Button("🧹 Clear History")
-    
-    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(
-        get_chat_response, [chatbot, task_history], [chatbot], show_progress=True
-    )
-    submit_btn.click(clear_input, [], [query])
-    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)
-
-# Launch the demo
-demo.launch()
+iface = gr.Interface(
+    fn=main_function,
+    inputs=[
+        gr.inputs.Textbox(lines=2, label='Input'),
+        gr.inputs.Image(type='file', label='Upload Image')
+    ],
+    outputs='text',
+    live=True,
+    layout='vertical',
+    theme='huggingface',
+    css=css
+)
+
+iface.add_button("🧹 Clear History", clear_history_fn)
+
+iface.launch(share=True)
+