Vertical Federated Learning Communicator

class mindspore_federated.VerticalFederatedCommunicator(http_server_config, remote_server_config, enable_ssl=False, ssl_config=None, compress_configs=None)[source]

Define the vertical federated communicator.

Parameters

  • http_server_config (ServerConfig) – Configuration of local http server.

  • remote_server_config (ServerConfig) – Configuration of remote http server.

  • enable_ssl (bool, optional) – whether to enable ssl communication. Default: False.

  • ssl_config (SSLConfig, optional) – Configuration of ssl encryption. Default: None.

  • compress_configs (dict, optional) – Configuration of communication compression. Default: None.

Examples

>>> from mindspore_federated import VerticalFederatedCommunicator, ServerConfig
>>> http_server_config = ServerConfig(server_name='server', server_address="127.0.0.1:1086")
>>> remote_server_config = ServerConfig(server_name='client', server_address="127.0.0.1:1087")
>>> vertical_communicator = VerticalFederatedCommunicator(http_server_config=http_server_config,
...                                                       remote_server_config=remote_server_config)
>>> vertical_communicator.launch()
>>> vertical_communicator.data_join_wait_for_start()
data_join_wait_for_start()[source]

Block and wait for the registration information of the client worker.

http_server_config()[source]

Returns the local server configuration.

launch()[source]

Start vertical federated learning communicator.

receive(target_server_name: str)[source]

Get the sensor data sent by the remote server.

Parameters

target_server_name (str) – Specifies the name of the remote server.

remote_server_config()[source]

Returns the remote server configuration.

send_tensors(target_server_name, tensor_dict)[source]

Send distributed training sensor data.

Parameters

  • target_server_name (str) – Specifies the name of the remote server.

  • tensor_dict (OrderedDict) – The dict of Tensors to be sent.

Examples

>>> from mindspore import Tensor
>>> backbone_out = OrderedDict()
>>> backbone_out["hidden_states"] = Tensor(np.random.random(size=(2, 2048, 1280)).astype(np.float16))
>>> vertical_communicator.send_tensors("leader", backbone_out)
class mindspore_federated.ServerConfig(server_name, server_address)[source]

Define the vertical server configuration.

Parameters

  • server_name (str) – Name of server, such as “leader_server”, user defined.

  • server_address (str) – Address of server, such as 127.0.0.1:1086, user defined.

class mindspore_federated.SSLConfig(server_password, client_password, server_cert_path=None, client_cert_path=None, ca_cert_path=None, crl_path=None, cipher_list=default_cipher_list, cert_expire_warning_time_in_day=90)[source]

Define the ssl certificate config. If you want to start SSL, you need to configure the following parameters. The return value is used for the third input of mindspore_federated.VerticalFederatedCommunicator .

Parameters

  • server_password (str) – The password of the server certificate. Default: None.

  • client_password (str) – The password of the client certificate. Default: None.

  • server_cert_path (str) – The absolute path of the server certificate on the server. Default: None.

  • client_cert_path (str) – The absolute path of the client certificate on the server. Default: None.

  • ca_cert_path (str) – The absolute path of the root certificate on the server. Default: None.

  • crl_path (str) – The absolute path of the CRL certificate on the server. Default: None.

  • cipher_list (str) – The server supports the default encryption suite for ssl communication. Default: default_cipher_list.

  • cert_expire_warning_time_in_day (int) – How many days before the certificate expires to start printing warning messages. Default: 90.

class mindspore_federated.CompressConfig(compress_type, bit_num=8)[source]

Define the vertical server compress config.

Parameters

  • compress_type (str) –

    Compress type for vertical fl communication. Supports [“min_max”, “bit_pack”].

    • min_max: The min max quantization compress method.

    • bit_pack: The bit pack compress method.

  • bit_num (int) – Bits num of quant algorithm. The value range is within [1, 8]. Default: 8.