openfl.component.envoy.envoy.Envoy

openfl.component.envoy.envoy.Envoy#

class openfl.component.envoy.envoy.Envoy(*, shard_name, director_host, director_port, shard_descriptor, root_certificate=None, private_key=None, certificate=None, tls=True, install_requirements=True, cuda_devices=(), cuda_device_monitor=None, review_plan_callback=None)[source]#

Bases: object

Envoy class. The Envoy is a long-lived entity that runs on collaborator nodes connected to the Director.

Parameters:

  • shard_name (str)

  • director_host (str)

  • director_port (int)

  • shard_descriptor (Type[ShardDescriptor])

  • root_certificate (Path | str | None)

  • private_key (Path | str | None)

  • certificate (Path | str | None)

  • tls (bool)

  • install_requirements (bool)

  • cuda_devices (tuple | list)

  • cuda_device_monitor (Type[CUDADeviceMonitor] | None)

  • review_plan_callback (None | Callable)

name#

The name of the shard.

Type:

str

root_certificate#

The path to the root certificate for TLS.

Type:

Union[Path, str]

private_key#

The path to the private key for TLS.

Type:

Union[Path, str]

certificate#

The path to the certificate for TLS.

Type:

Union[Path, str]

director_client#

The director client.

Type:

ShardDirectorClient

shard_descriptor#

The shard descriptor.

Type:

Type[ShardDescriptor]

cuda_devices#

The CUDA devices.

Type:

tuple

install_requirements#

A flag indicating if the requirements should be installed.

Type:

bool

review_plan_callback#

A callback function for reviewing the plan.

Type:

Union[None, Callable]

cuda_device_monitor#

The CUDA device monitor.

Type:

Optional[Type[CUDADeviceMonitor]]

executor#

The executor for running tasks.

Type:

ThreadPoolExecutor

running_experiments#

A dictionary to store the running experiments.

Type:

dict

is_experiment_running#

A flag indicating if an experiment is running.

Type:

bool

_health_check_future#

The future object for the health check.

Type:

object

__init__(*, shard_name, director_host, director_port, shard_descriptor, root_certificate=None, private_key=None, certificate=None, tls=True, install_requirements=True, cuda_devices=(), cuda_device_monitor=None, review_plan_callback=None)[source]#

Initialize a envoy object.

Parameters:

  • shard_name (str) – The name of the shard.

  • director_host (str) – The host of the director.

  • director_port (int) – The port of the director.

  • shard_descriptor (Type[ShardDescriptor]) – The shard descriptor.

  • root_certificate (Optional[Union[Path, str]], optional) – The path to the root certificate for TLS. Defaults to None.

  • private_key (Optional[Union[Path, str]], optional) – The path to the private key for TLS. Defaults to None.

  • certificate (Optional[Union[Path, str]], optional) – The path to the certificate for TLS. Defaults to None.

  • tls (bool, optional) – A flag indicating if TLS should be used for connections. Defaults to True.

  • install_requirements (bool, optional) – A flag indicating if the requirements should be installed. Defaults to True.

  • cuda_devices (Union[tuple, list], optional) – The CUDA devices. Defaults to ().

  • cuda_device_monitor (Optional[Type[CUDADeviceMonitor]], optional) – The CUDA device monitor. Defaults to None.

  • review_plan_callback (Union[None, Callable], optional) – A callback function for reviewing the plan. Defaults to None.

Return type:

None

Methods

__init__(*, shard_name, director_host, ...)

Initialize a envoy object.

run()

Run of the envoy working cycle.

send_health_check()

Send health check to the director.

start()

Start the envoy.
run()[source]#

Run of the envoy working cycle.

send_health_check()[source]#

Send health check to the director.

start()[source]#

Start the envoy.