nvflare.private.fed.client.client_executor module

class ClientExecutor[source]

Bases: ABC

abstract abort_app(job_id)[source]

Aborts the running app.

Parameters:

job_id – the job_id

abstract abort_task(job_id)[source]

Aborts the client executing task.

Parameters:

job_id – the job_id

abstract check_status(job_id) → str[source]

Checks the status of the running client.

Parameters:

job_id – the job_id

Returns:

A client status message

abstract get_errors(job_id)[source]

Get the error information.

Returns:

A dict of error information.

abstract get_run_info(job_id)[source]

Gets the run information.

Parameters:

job_id – the job_id

Returns:

A dict of run information.

abstract reset_errors(job_id)[source]

Resets the error information.

Parameters:

job_id – the job_id

abstract start_app(client, job_id, job_meta, args, allocated_resource, token, resource_manager, fl_ctx: FLContext)[source]

Starts the client app.

Parameters:

  • client – the FL client object

  • job_id – the job_id

  • args – admin command arguments for starting the FL client training

  • allocated_resource – allocated resources

  • token – token from resource manager

  • resource_manager – resource manager

  • fl_ctx – FLContext

class JobExecutor(client, startup)[source]

Bases: ClientExecutor

Run the Client executor in a child process.

To init the ProcessExecutor.

Parameters:

startup – startup folder

abort_app(job_id)[source]

Aborts the running app.

Parameters:

job_id – the job_id

abort_task(job_id)[source]

Aborts the client executing task.

Parameters:

job_id – the job_id

check_status(job_id)[source]

Checks the status of the running client.

Parameters:

job_id – the job_id

Returns:

A client status message

configure_job_log(job_id, config)[source]

Configure the job log.

Parameters:

  • job_id – the job_id

  • config –

    log config

    Returns:

    configure_job_log command message

get_errors(job_id)[source]

Get the error information.

Parameters:

job_id – the job_id

Returns:

A dict of error information.

get_run_info(job_id)[source]

Gets the run information.

Parameters:

job_id – the job_id

Returns:

A dict of run information.

get_run_processes_keys()[source]
get_status(job_id)[source]
notify_job_status(job_id, job_status)[source]
reset_errors(job_id)[source]

Resets the error information.

Parameters:

job_id – the job_id

send_to_job(job_id, channel: str, topic: str, msg: Message, timeout: float, optional=False) → Message[source]

Send a message to CJ

Parameters:

  • job_id – id of the job

  • channel – message channel

  • topic – message topic

  • msg – the message to be sent

  • timeout – how long to wait for reply

  • optional – whether the message is optional

Returns: reply from CJ

start_app(client, job_id, job_meta, args, allocated_resource, token, resource_manager: ResourceManagerSpec, fl_ctx: FLContext)[source]

Starts the app.

Parameters:

  • client – the FL client object

  • job_id – the job_id

  • job_meta – job metadata

  • args – admin command arguments for starting the worker process

  • allocated_resource – allocated resources

  • token – token from resource manager

  • resource_manager – resource manager

  • fl_ctx – FLContext