app.py

import gradio as gr
from threading import Thread
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
import torch
import os

# --- 配置 ---
MODEL_ID = "Qwen/Qwen3-0.6B-Base"
# ❗ 1. 定义系统提示词
SYSTEM_PROMPT = """As a helper/no_think
"""
# --- 加载模型和分词器 ---
print("开始加载模型和分词器...")
try:
    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
    model = AutoModelForCausalLM.from_pretrained(
        MODEL_ID,
        torch_dtype="auto",
        device_map="auto",
        trust_remote_code=True
    )
    print("模型和分词器加载成功！")
except Exception as e:
    print(f"模型加载失败: {e}")
    raise gr.Error(f"关键错误：无法加载模型 {MODEL_ID}。错误信息: {e}")

# --- 核心对话函数 ---
def predict(message, history):
    messages = []
    
    # ❗ 2. 在消息列表的开头添加系统提示词
    messages.append({"role": "system", "content": SYSTEM_PROMPT})
    
    # # 3. 添加历史对话记录
    # for turn in history:
    #     user_msg, assistant_msg = turn
    #     messages.append({"role": "user", "content": user_msg})
    #     messages.append({"role": "assistant", "content": assistant_msg})
        
    # 4. 添加当前的用户消息
    messages.append({"role": "user", "content": message})
    
    # 使用 tokenizer.apply_chat_template 将消息列表转换为模型输入
    model_inputs = tokenizer.apply_chat_template(
        messages, 
        add_generation_prompt=True, 
        return_tensors="pt"
    ).to(model.device)

    streamer = TextIteratorStreamer(tokenizer, timeout=300.0, skip_prompt=True, skip_special_tokens=True)

    generation_kwargs = dict(
        inputs=model_inputs,
        streamer=streamer,
        max_new_tokens=2048,
        do_sample=True,
        temperature=0.4,
        top_p=0.95,
    )
    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()

    full_response = ""
    for new_text in streamer:
        full_response += new_text
        yield full_response

# --- 创建并启动Gradio界面 ---
# 已移除 examples 和 cache_examples 参数来修复点击示例时报错的问题
demo = gr.ChatInterface(
    fn=predict
)

if __name__ == "__main__":
    # 使用 share=True 来允许跨域 WebSocket 连接
    demo.queue().launch(share=True)