Recursively copy this CfgNode.

load configs from another cfg instance Args: cfg_other: other cfg to be merged check_cfg: whether enable ``a

Perform ML split and FL split. Returns: dict of ``ClientData`` with client_idx as key to build \ ``Standalon

To bind a message type with a handling function. Arguments: msg_type (str): The defined message type callbac

Instantiate the data and update the configuration accordingly if necessary. Arguments: config: a cfg node object client_cfgs:

To run an FL course, which is called after server/client has been set up. For the standalone mode, a shared message queue wil

This function return a class of client. Args: cfg: configurations for FL, see ``federatedscope.core.configs`` Returns:

This function return a class of server. Args: cfg: configurations for FL, see ``federatedscope.core.configs`` Returns:

Instantiate a runner based on a configuration file Args: server_class: server class client_class: client class confi

load configs from a list stores the keys and values. \ modified ``merge_from_list`` in ``yacs.config.py`` to allow adding \ n

Called by the FL client to update the model parameters Arguments: model_parameters (dict): {model_name: model_val}

To modify the state of clients (idle or working)

Dump to a string.

Format the evaluation results from ``trainer.ctx.eval_results`` Args: results (dict): a dict to store the evaluation res

To conduct evaluation. When ``cfg.federate.make_global_eval=True``, \ a global evaluation is conducted by the server.

load configs from a yaml file, another cfg instance or a list \ stores the keys and values. Args: cfg_filename:

(1) make the cfg attributes immutable; (2) if ``save==True``, save the frozen cfg_check_funcs into \ ``self.outdir/config

copy the refs.

This function builds an instance of model to be trained. Arguments: config: ``cfg`` local_data: the model to be instantiated

Downloads the content of an url to a folder. Modified from \ https://github.com/pyg-team/pytorch_geometric/tree/master/torch_geometric A

Make this CfgNode and all of its children mutable.

To check the message buffer Arguments: cur_round (int): The current round number min_received_num (int): The

Arguments: arg_opts: list pairs of arg.opts

Register a function that checks the configuration node. Args: cfg_check_fun: function for validation the correctness of

This function returns an instantiated optimizer to optimize the model. Args: model: model to be optimized type: type of opti

returns trace of one of three tracked quantities Args: key (str): 'entropy', 'global', or 'refine' Returns: nu

Update best evaluation results. \ by default, the update is based on validation loss with \ ``round_wise_update_key="val_loss

To broadcast the message to all clients or sampled clients Arguments: msg_type: 'model_para' or other user defined msg_t

Merge two ``dict`` according to their keys, and concatenate their value. Args: dict1: ``dict`` to be merged dict2: ``dict``

Save formatted results to a file.

Checks the new result and if it improves it returns True. Args: new_result: new evaluation result Returns:

This function is to build transforms applying to dataset. Args: config: ``CN`` from ``federatedscope/core/configs/config.py``

Args: data: ``federatedscope.core.data.ClientData`` with Tabular format. Returns: Merged data feature/x.

samples from configs using current probability vector Arguments: thetas (list): probabilities for the hyperparameters.

To simulate the message handling process (used only for the \ standalone mode)

To simulate the message handling process (used only for the standalone mode)

model parameter filter when transmit between local and gloabl, which is useful in personalization. e.g., setting cfg.personal

Aggregation function. Args: agg_info: information to be aggregated.

To allocate a device

To broadcast the communication addresses of clients (used for \ additive secret sharing)

To check the message_buffer. When enough messages are receiving, \ some events (such as perform aggregation, evaluation, and move to

To save the results and save model after each evaluation, and check \ whether to early stop.

r""" Returns: the split edge index.

Increment the model weight by the given content.

To send ``join_in`` message to the server for joining in the FL course.

Merge evaluation results from all clients, update best, \ log the merged results and save them into eval_results.log Returns