Update app.py

app.py

@@ -1,52 +1,100 @@
 from huggingface_hub import InferenceClient
 import gradio as gr
+import datetime
 
 client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 
 def format_prompt(message, history):
-  prompt = "<s>"
-  for user_prompt, bot_response in history:
-    prompt += f"[INST] {user_prompt} [/INST]"
-    prompt += f" {bot_response}</s> "
-  prompt += f"[INST] {message} [/INST]"
-  return prompt
+    prompt = "<s>"
+    for user_prompt, bot_response in history:
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
+    prompt += f"[INST] {message} [/INST]"
+    return prompt
 
 def generate(
-    prompt, history, temperature=0.2, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
+    message, history, temperature=0.2, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
 ):
-    temperature = float(temperature)
-    if temperature < 1e-2:
-        temperature = 1e-2
+    temperature = max(float(temperature), 1e-2)
     top_p = float(top_p)
 
     generate_kwargs = dict(
         temperature=temperature,
-        max_new_tokens=max_new_tokens,
+        max_new_tokens=int(max_new_tokens),
         top_p=top_p,
         repetition_penalty=repetition_penalty,
         do_sample=True,
         seed=42,
     )
 
-    formatted_prompt = format_prompt(prompt, history)
+    formatted_prompt = format_prompt(message, history)
 
-    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
+    # Логирование промпта
+    with open("conversation_log.txt", "a", encoding="utf-8") as f:
+        f.write(f"{datetime.datetime.now()}\n")
+        f.write(f"Промпт: {formatted_prompt}\n")
+
+    stream = client.text_generation(
+        formatted_prompt,
+        **generate_kwargs,
+        stream=True,
+        details=True,
+        return_full_text=False,
+    )
     output = ""
 
     for response in stream:
         output += response.token.text
         yield output
-    return output
 
-    
-mychatbot = gr.Chatbot(
-    avatar_images=["./user.png", "./botm.png"], bubble_full_width=False, show_label=False, show_copy_button=True, likeable=True,)
+    # Логирование ответа
+    with open("conversation_log.txt", "a", encoding="utf-8") as f:
+        f.write(f"Ответ: {output}\n\n")
+
+    # Обновление истории
+    history.append((message, output))
+
+def update_history(instruction, model_answer):
+    history = []
+    if instruction and model_answer:
+        history.append((instruction, model_answer))
+    return history
 
-demo = gr.ChatInterface(fn=generate, 
-                        chatbot=mychatbot,
-                        #title="WebpyGPT",
-                        retry_btn=None,
-                        undo_btn=None
-                       )
+with gr.Blocks() as demo:
+    gr.Markdown("# Чат с Mixtral-8x7B-Instruct-v0.1")
+
+    instruction = gr.Textbox(label="Instruction", placeholder="Введите начальную инструкцию")
+    model_answer = gr.Textbox(label="Model Answer", placeholder="Введите ответ модели на инструкцию")
+    set_initial_btn = gr.Button("Установить начальный диалог")
+
+    chatbot = gr.Chatbot(
+        avatar_images=["./user.png", "./botm.png"],
+        bubble_full_width=False,
+        show_label=False,
+        show_copy_button=True,
+        likeable=True,
+    )
+    follow_up_instruction = gr.Textbox(label="Follow-up Instruction", placeholder="Введите ваше сообщение")
+
+    history_state = gr.State([])
+
+    set_initial_btn.click(
+        fn=update_history,
+        inputs=[instruction, model_answer],
+        outputs=[history_state],
+    )
+
+    def respond(message, history):
+        gen = generate(message, history)
+        response = ""
+        for res in gen:
+            response = res
+        return history + [(message, response)], history + [(message, response)]
+
+    follow_up_instruction.submit(
+        fn=respond,
+        inputs=[follow_up_instruction, history_state],
+        outputs=[chatbot, history_state],
+    )
 
-demo.queue().launch(show_api=False)
+demo.queue().launch(show_api=False)