app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# 加载DeepSeek-Coder-6.7B模型
model_name = "deepseek/DeepSeek-Coder-6.7B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,  # 使用半精度以减少显存
    device_map="auto",         # 自动分配到GPU
    trust_remote_code=True     # DeepSeek模型可能需要
)

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    # 格式化输入
    input_text = tokenizer.apply_chat_template(messages, tokenize=False)
    inputs = tokenizer(input_text, return_tensors="pt").to("cuda")

    # 生成响应
    outputs = model.generate(
        **inputs,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
    )
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    yield response

# Gradio界面
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly coding assistant.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)

if __name__ == "__main__":
    demo.launch()