pickle_dataset

本数据集是由用户自行创建并上传至 ModelScope 平台的测试数据集，主要用于测试自定义数据加载与模型行为，特别是用于验证 Pickle 格式在加载过程中可能引发的安全问题（例如反序列化攻击）。**该数据集仅用于个人测试用途，**不建议在生产环境中使用。

数据集中包含一个或多个以 .pkl 结尾的文件，这些文件是通过 Python 的 pickle 模块序列化的对象。在使用过程中请确保数据来源可信，否则可能存在安全风险。

下载方式
你可以使用 ModelScope SDK 提供的命令行工具来下载该数据集。支持通过 modelscope dataset 命令快速拉取数据集文件，默认下载到本地缓存目录，或通过 --local_dir 参数指定保存位置。

命令行方式下载
```
modelscope download --dataset jiruwozhi/pickle_dataset
```
你也可以通过指定输出目录进行保存：
```
modelscope download --dataset jiruwozhi/pickle_dataset --local_dir ./my_dataset
```
或者仅下载部分文件（例如只下载某个 sample.pkl 文件）：
```
modelscope download --dataset jiruwozhi/pickle_dataset sample.pkl
```
若数据集为私有状态，需通过 --token 参数提供访问令牌。

数据使用说明
数据结构：包含若干 .pkl 文件

主要用途：Pickle 文件加载安全性验证、反序列化测试、ModelScope 自定义数据加载测试

注意事项：请勿在不可信来源环境中使用 pickle 文件

This is a test dataset self-created by users and uploaded to the ModelScope platform. It is primarily used to test custom data loading and model behavior, particularly to verify potential security issues (e.g., deserialization attacks) that may arise during the loading of Pickle format files. **This dataset is for personal testing purposes only** and is not recommended for production environments.

This dataset contains one or more files ending with the .pkl suffix, which are objects serialized via Python's pickle module. Please ensure the data source is trusted during usage, otherwise security risks may exist.

### Download Methods
You can use the command-line tool provided by the ModelScope SDK to download this dataset. It supports quickly fetching dataset files via the `modelscope dataset` command, with files downloaded to the local cache directory by default, or you can specify the save location via the `--local_dir` parameter.

#### Command-line Download Examples

modelscope download --dataset jiruwozhi/pickle_dataset

You can also specify an output directory to save the files:

modelscope download --dataset jiruwozhi/pickle_dataset --local_dir ./my_dataset

Or download only specific files (e.g., only download the sample.pkl file):

modelscope download --dataset jiruwozhi/pickle_dataset sample.pkl

If the dataset is in private status, you need to provide an access token via the `--token` parameter.

### Data Usage Instructions
**Data Structure**: Contains multiple .pkl files
**Primary Usage**: Security verification of Pickle file loading, deserialization testing, and custom data loading testing on ModelScope
**Notes**: Do not use pickle files in untrusted source environments