flytekitplugins.kftensorflow.TfJob#

class flytekitplugins.kftensorflow.TfJob(chief=<factory>, ps=<factory>, worker=<factory>, evaluator=<factory>, run_policy=<factory>, num_workers=None, num_ps_replicas=None, num_chief_replicas=None, num_evaluator_replicas=None)[source]#

Configuration for an executable TensorFlow Job. Use this to run distributed TensorFlow training on Kubernetes.

Parameters

  • chief (flytekitplugins.kftensorflow.task.Chief) – Configuration for the chief replica group.

  • ps (flytekitplugins.kftensorflow.task.PS) – Configuration for the parameter server (PS) replica group.

  • worker (flytekitplugins.kftensorflow.task.Worker) – Configuration for the worker replica group.

  • evaluator (flytekitplugins.kftensorflow.task.Evaluator) – Configuration for the evaluator replica group.

  • run_policy (Optional[flytekitplugins.kftensorflow.task.RunPolicy]) – Configuration for the run policy.

  • num_workers (Optional[int]) – [DEPRECATED] This argument is deprecated. Use worker.replicas instead.

  • num_ps_replicas (Optional[int]) – [DEPRECATED] This argument is deprecated. Use ps.replicas instead.

  • num_chief_replicas (Optional[int]) – [DEPRECATED] This argument is deprecated. Use chief.replicas instead.

  • num_evaluator_replicas (Optional[int]) –

Return type

None

Methods

Attributes

num_chief_replicas: Optional[int] = None
num_evaluator_replicas: Optional[int] = None
num_ps_replicas: Optional[int] = None
num_workers: Optional[int] = None
chief: flytekitplugins.kftensorflow.task.Chief
ps: flytekitplugins.kftensorflow.task.PS
worker: flytekitplugins.kftensorflow.task.Worker
evaluator: flytekitplugins.kftensorflow.task.Evaluator
run_policy: Optional[flytekitplugins.kftensorflow.task.RunPolicy]