kubeflow.fairing.ml_tasks package

Submodules

kubeflow.fairing.ml_tasks.tasks module

class kubeflow.fairing.ml_tasks.tasks.BaseTask(entry_point, base_docker_image=None, docker_registry=None, input_files=None, backend=None, pod_spec_mutators=None)

Bases: object

Base class for handling high level ML tasks.

Parameters:
  • entry_point – An object or reference to the source code that has to be deployed.
  • base_docker_image – Name of the base docker image that should be used as a base image when building a new docker image as part of an ML task deployment.
  • docker_registry – Docker registry to store output docker images.
  • input_files – list of files that needs to be packaged along with the entry point. E.g. local python modules, trained model weigths, etc.
class kubeflow.fairing.ml_tasks.tasks.PredictionEndpoint(model_class, base_docker_image=None, docker_registry=None, input_files=None, backend=None, service_type='ClusterIP', pod_spec_mutators=None)

Bases: kubeflow.fairing.ml_tasks.tasks.BaseTask

Create a prediction endpoint.

create()

Create prediction endpoint.

delete()

Delete prediction endpoint.

predict_nparray(data, feature_names=None)

Return the prediction result.

Parameters:
  • data – Data to be predicted.
  • feature_names – Feature extracted from data (Default value = None)
class kubeflow.fairing.ml_tasks.tasks.TrainJob(entry_point, base_docker_image=None, docker_registry=None, input_files=None, backend=None, pod_spec_mutators=None)

Bases: kubeflow.fairing.ml_tasks.tasks.BaseTask

Create a train job.

submit()

Submit a train job.

kubeflow.fairing.ml_tasks.utils module

kubeflow.fairing.ml_tasks.utils.guess_preprocessor(entry_point, input_files, output_map)

Preprocessor to use to modify inputs before sending them to docker build

Parameters:
  • entry_point – entry_point which to use
  • input_files – input files
  • output_map – output
kubeflow.fairing.ml_tasks.utils.is_docker_daemon_exists()

To check if docker daemon exists or not.

Module contents