kfp.components¶

The kfp.components module contains functions for loading components from compiled YAML.

Functions:

Classes:

kfp.components.load_component_from_file(file_path: str) → YamlComponent[source]¶

Loads a component from a file.

Parameters¶

file_path : str¶

Filepath to a YAML component.

Returns¶

Component loaded from YAML.

Example

from kfp import components

components.load_component_from_file('~/path/to/pipeline.yaml')

kfp.components.load_component_from_url(url: str, auth: tuple[str, str] | None = None) → YamlComponent[source]¶

Loads a component from a URL.

Parameters¶

url : str¶

URL to a YAML component.

auth : Tuple[str, str], optional¶

A ('<username>', '<password>') tuple of authentication credentials necessary for URL access. See Requests Authorization for more information.

Returns¶

Component loaded from YAML.

Example

from kfp import components

components.load_component_from_url('https://raw.githubusercontent.com/kubeflow/pipelines/7b49eadf621a9054e1f1315c86f95fb8cf8c17c3/sdk/python/kfp/compiler/test_data/components/identity.yaml')

components.load_component_from_url('gs://path/to/pipeline.yaml')

kfp.components.load_component_from_text(text: str) → YamlComponent[source]¶

Loads a component from text.

Parameters¶

text : str¶

Component YAML text.

Returns¶

Component loaded from YAML.

class kfp.components.PythonComponent(component_spec: ComponentSpec, python_func: Callable)[source]¶

Bases: BaseComponent

A component defined via Python function.

Note: PythonComponent is not intended to be used to construct components directly. Use @kfp.dsl.component instead.

Parameters¶

component_spec: ComponentSpec¶

Component definition.

python_func: Callable¶

Python function that becomes the implementation of this component.

Methods:

execute(**kwargs)

Executes the Python function that defines the component.

execute(**kwargs)[source]¶

Executes the Python function that defines the component.

class kfp.components.BaseComponent(component_spec: ComponentSpec)[source]¶

Bases: ABC

Base class for a component.

Note: BaseComponent is not intended to be used to construct components directly. Use @kfp.dsl.component or kfp.components.load_component_from_*() instead.

name¶

Name of the component.

component_spec¶

Component definition.

Attributes:

pipeline_spec	Returns the pipeline spec of the component.
platform_spec	Returns the PlatformSpec of the component.
required_inputs

Methods:

execute(**kwargs)

Executes the component locally if implemented by the inheriting subclass.

property pipeline_spec : PipelineSpec¶

Returns the pipeline spec of the component.

property platform_spec : PlatformSpec¶

Returns the PlatformSpec of the component.

Useful when the component is a GraphComponent, else will be empty per component_spec.platform_spec default.

abstract execute(**kwargs)[source]¶

Executes the component locally if implemented by the inheriting subclass.

property required_inputs : list[str]¶

class kfp.components.ContainerComponent(component_spec: ComponentSpec, pipeline_func: Callable)[source]¶

Bases: BaseComponent

Component defined via pre-built container.

Attribute:

pipeline_func: The function that becomes the implementation of this component.

Methods:

execute(**kwargs)

Executes the component locally if implemented by the inheriting subclass.

execute(**kwargs)[source]¶

Executes the component locally if implemented by the inheriting subclass.

class kfp.components.YamlComponent(component_spec: ComponentSpec, component_yaml: str)[source]¶

Bases: BaseComponent

A component loaded from a YAML file.

Note: YamlComponent is not intended to be used to construct components directly. Use kfp.components.load_component_from_*() instead.

Attribute:

component_spec: Component definition. component_yaml: The yaml string that this component is loaded from.

Attributes:

pipeline_spec

Returns the pipeline spec of the component.

Methods:

execute(*args, **kwargs)

Not implemented.

property pipeline_spec : PipelineSpec¶

Returns the pipeline spec of the component.

execute(*args, **kwargs)[source]¶

Not implemented.