This function builds an instance of loss functions from: \ "https://pytorch.org/docs/stable/nn.html#loss-functions", where the ``criterio

This function builds an instance of regularizer to regularize training. Args: reg_type: type of scheduler, such as see \

This function is to build splitter to generate simulated federation \ datasets from non-FL dataset. Args: config: configurations

Calculate wall time and round when global convergence has been reached.

r""" Returns: calculate the hamming distance of graph data

Query the benchmark using a given configuration and a ( round, sample_client_rate) (=budget). Parameters ----------

print some meta info for code-users, e.g., model type; the para names will be filtered out, etc.,

Check and cleans up the internal state of cfg and save cfg. Args: check_cfg: whether enable ``assert_cfg()``

return the img: np.array [0:255], (height, width, channel)

Set up ``DataLoader`` in ``ClientData`` with new configurations. Args: new_client_cfg: new client-specific CfgNode

Get a tokenizer, the default bos/eos/eoq token is used for Bert

Args: feat: feature to be binned, which must be 2D numpy.array num_bins: list for bins strategy: binning strategy, ``'un

Register noise injector for the server

To generate a graph with specified homophilic degree, avg node degree, feature dimension, etc. Arguments: n (int): the number of nodes

Helper method that applies process_func to the files in the system and model folders and saves the resulting files to new sys

[Modified from yacs, to allow int <-> float conversation] Checks that ``replacement``, which is intended to replace \ ``original`` is of

Decodes a raw config value (e.g., from a yaml config files or command \ line argument) into a Python object. (1) If the valu

Source: https://github.com/eric-mitchell/direct-preference-optimization/ blob/main/trainers.py#L208 Compute the log probabilities of

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

Set immutability to is_immutable and recursively apply the setting \ to all nested CfgNodes.

Load a config from a YAML string encoding.

[Modified from yacs, to allow int <-> float conversation] Merge config dictionary a into config dictionary b, clobbering the \ options i

Run local evaluation, return some metric to maximize (e.g. val_acc)

Register default handler dic to handle message, which includes \ sender, receiver, state, and content. More detail can be found in \

Populate ``${split}_data``, ``${split}_loader`` and \ ``num_${split}_data`` for different data splits, and setup init var \ i

Add a new key: value item (results-type: new_results) to best_result

Args: parameter_updates: Tensor with dimension n * d_feature prop: Tensor with dimension n * 1 Returns:

Arguments: last_model (dict): the state of last round. local_updated_models (list): each element is (data_size, model). Retu

To save the results and save model after each evaluation