app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

# 加载模型（使用量化版本节省显存）
model_name = "Qwen/Qwen2.5-7B-Instruct-GPTQ-Int8"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="auto",  # 自动分配GPU/CPU
    torch_dtype="auto"
)

# 定义生成函数
def generate_response(message, history):
    # 格式化对话历史
    messages = [{"role": "user", "content": message}]
    text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    inputs = tokenizer(text, return_tensors="pt").to(model.device)
    
    # 生成回复
    outputs = model.generate(
        **inputs,
        max_new_tokens=512,
        temperature=0.7
    )
    response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
    return response

# 启动Gradio界面
gr.ChatInterface(
    fn=generate_response,
    title="Qwen2.5-7B大模型在线演示",
    description="输入问题后按回车开始对话"
).launch()