update

app.py

@@ -1,21 +1,4 @@
 """
-来自 https://github.com/OpenLMLab/MOSS/blob/main/moss_web_demo_gradio.py
-
-
-# 难点
-
-## TODO
-
-
--[x] 代码和表格的预览
--[x] markdown解析：mdtex2html
--[ ] 可编辑chatbot：https://github.com/gradio-app/gradio/issues/4444
--[ ] 乱码问题
-
-
-## Reference
-
-- https://github.com/GaiZhenbiao/ChuanhuChatGPT/
 """
 
 import config
@@ -29,51 +12,61 @@ system_list = [
     "你是一个心理咨询师。",
 ]
 
+
+doc = """\
+There are maily two types of user simulator:
+- prompt-based user-simulator (role-play)
+- model-based user-simulator
 """
-TODO: 使用说明
-"""
+
 with gr.Blocks() as demo:
     # Knowledge Distillation through Self Chatting
     gr.HTML("""<h1 align="center">Distilling the Knowledge through Self Chatting</h1>""")
+    gr.Markdown(doc, visible=False)
     with gr.Row():
-        system = gr.Dropdown(
-            choices=system_list,
-            value=system_list[0],
-            allow_custom_value=True,
-            interactive=True,
-            label="System message",
-            scale=4,
-        )
         gr.Dropdown(
             ["Qwen2-0.5B-Instruct", "llama3.1", "gemini"],
             value="Qwen2-0.5B-Instruct",
             label="model",
             interactive=True,
             scale=1,
+            visible=False
+        )
+        system = gr.Dropdown(
+            choices=system_list,
+            value=system_list[0],
+            allow_custom_value=True,
+            interactive=True,
+            label="System message",
+            scale=5,
         )
+
     chatbot = gr.Chatbot(avatar_images=("assets/man.png", "assets/bot.png"))
     with gr.Row():
+        with gr.Column():
+            slider_max_tokens = gr.Slider(minimum=1, maximum=config.MAX_SEQUENCE_LENGTH,
+                                          value=config.DEFAULT_MAX_TOKENS, step=1, label="Max tokens")
+            slider_temperature = gr.Slider(minimum=0.1, maximum=10.0,
+                                           value=config.DEFAULT_TEMPERATURE, step=0.1, label="Temperature",
+                                           info="Larger temperature increase the randomness")
+            slider_top_p = gr.Slider(
+                minimum=0.1,
+                maximum=1.0,
+                value=config.DEFAULT_TOP_P,
+                step=0.05,
+                label="Top-p (nucleus sampling)",
+            )
+
         with gr.Column(scale=4):
             generated_text = gr.Textbox(show_label=False, placeholder="...", lines=10, visible=False)
+            generate_btn = gr.Button("🤔️ Generate", variant="primary")
             with gr.Row():
-                generate_btn = gr.Button("🤔️ Generate")
                 retry_btn = gr.Button("🔄  Regenerate")
                 undo_btn = gr.Button("↩️ Undo")
                 clear_btn = gr.Button("🗑️  Clear")  # 🧹 Clear History (清除历史)
-                stop_btn = gr.Button("停止生成", variant="primary")
+                # stop_btn = gr.Button("停止生成", variant="primary", visible=False)
+
 
-        slider_max_tokens = gr.Slider(minimum=1, maximum=config.MAX_SEQUENCE_LENGTH,
-                                      value=config.DEFAULT_MAX_TOKENS, step=1, label="Max tokens")
-        slider_temperature = gr.Slider(minimum=0.1, maximum=10.0,
-                                       value=config.DEFAULT_TEMPERATURE, step=0.1, label="Temperature",
-                                       info="Larger temperature increase the randomness")
-        slider_top_p = gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=config.DEFAULT_TOP_P,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        )
 
     ########
     history = gr.State([{"role": "system", "content": system_list[0]}])

app_util.py

@@ -1,7 +1,8 @@
 import json
 import gradio as gr
 from utils.logging_util import logger
-from models.cpp_qwen2 import bot
+# from models.cpp_qwen2 import bot
+from models.hf_qwen2 import bot
 
 
 #