Can not pickle `torch.utils.data.DataLoader`

首先复现使用 torch.utils.data.DataLoader 的 cloudpickle.dumps 调用，重点关注 torch.utils.data.dataloader.py 中显示的 getstate 失败。该 issue 没有指出 cloudpickle 文件或测试，因此首先确定预期的序列化行为，并在实现任何内容之前为受支持的情况定义测试。

Hi team,

I'm working on some distributed system for AI, and I'd like to pickle an object to transfer it. But it seems I can not pickle DataLoader, is there any suggestions or best practice for that?

  File "/opt/examples/ps/.venv/lib/python3.12/site-packages/flamepy/client.py", line 325, in create_session
    object_ref = put_object(session_id, attrs.common_data)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/examples/ps/.venv/lib/python3.12/site-packages/flamepy/cache.py", line 74, in put_object
    data = cloudpickle.dumps(obj, protocol=cloudpickle.DEFAULT_PROTOCOL)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/examples/ps/.venv/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1544, in dumps
    cp.dump(obj)
  File "/opt/examples/ps/.venv/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1313, in dump
    return super().dump(obj)
           ^^^^^^^^^^^^^^^^^
  File "/opt/examples/ps/.venv/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 761, in __getstate__
    raise NotImplementedError("{} cannot be pickled", self.__class__.__name__)
NotImplementedError: ('{} cannot be pickled', '_SingleProcessDataLoaderIter')