【无标题】

 在配置文件处将host节点扩充至10个

  guest:
    - '9999'
  host:
    - '10000'
    - '10001'
    - '10002'
    - '10003'
    - '10004'
    - '10005'
    - '10006'
    - '10007'
    - '10008'
    - '10009'
  arbiter:
    - '10000'

 假设每个节点都使用一份数据，便可以使用for循环进行简单控制，上传乳腺癌诊断的横向联邦学习的数据集

for i in range(10):
    pipeline.transform_local_file_to_dataframe(   file="/data/projects/fate/examples/data/breast_homo_host.csv",
        meta=meta, head=True, extend_sid=True,
        namespace="experiment",
        name=f"breast_homo_host_{i}")

 循环使用不同的host节点id初始化pipeline

for i in range(10):
        host = parties.host[i]
        pipeline = FateFlowPipeline().set_parties(guest=guest, host=host, arbiter=arbiter)

 训练完成

 设置训练配置的方法源码

def get_config_of_default_runner(
    algo: str = "fedavg",
    model: Union[TorchModule, Sequential, ModelLoader] = None,
    optimizer: Union[TorchOptimizer, Loader] = None,
    loss: Union[TorchModule, CustFuncLoader] = None,
    training_args: TrainingArguments = None,
    fed_args: FedArguments = None,
    dataset: DatasetLoader = None,
    data_collator: CustFuncLoader = None,
    tokenizer: CustFuncLoader = None,
    task_type: Literal["binary", "multi", "regression", "causal_lm", "others"] = "binary",
):

    if model is not None and not isinstance(
        model, (TorchModule, Sequential, ModelLoader)
    ):
        raise ValueError(
            f"The model is of type {type(model)}, not TorchModule, Sequential, or ModelLoader. Remember to use patched_torch_hook for passing NN Modules or Optimizers."
        )

    if fed_args is not None and not isinstance(fed_args, FedArguments):
        raise ValueError(
            f"Federation arguments are of type {type(fed_args)}, not FedArguments."
        )

    runner_conf = _get_config_of_default_runner(
        optimizer, loss, training_args, dataset, data_collator, tokenizer, task_type
    )
    runner_conf['algo'] = algo
    runner_conf['model_conf'] = model.to_dict() if model is not None else None
    runner_conf['fed_args_conf'] = fed_args.to_dict() if fed_args is not None else None

    return runner_conf

model项可以是TorchModule，Sequential，ModelLoader的实例，表示模型

• 对于 model 参数，如果它不为 None 并且不是 TorchModule、Sequential 或 ModelLoader 的实例，则抛出一个 ValueError。
• 对于 fed_args 参数，如果它不为 None 并且不是 FedArguments 的实例，则也抛出一个 ValueError。
• 方法返回填充了所有必要信息的 runner_conf 字典。

from torch.nn import Sequential as tSequential
class Sequential(tSequential):
    def to_dict(self):
        """
        get the structure of current sequential
        """
        layer_confs = {}
        idx = 0
        for k in self._modules:
            ordered_name = idx
            layer_confs[ordered_name] = self._modules[k].to_dict()
            idx += 1
        ret_dict = {
            "module_name": "fate.components.components.nn.torch.base",
            "item_name": load_seq.__name__,
            "kwargs": {"seq_conf": layer_confs},
        }
        return ret_dict

    def to_json(self):
        return json.dumps(self.to_dict(), indent=4)