Reusable Kubeflow Components for ppc64le

Reusable Kubeflow components for ppc64le (IBM Power processor architecture).

Using Components

Kubeflow Pipelines Python SDK

import kfp
import kfp.dsl as dsl
from kfp import components

example_comp = components.load_component_from_url('https://raw.githubusercontent.com/lehrig/kubeflow-ppc64le-components/main/examples/hello-world/component.yaml')

@dsl.pipeline(
    name='My first pipeline',
    description='A hello world pipeline.'
)
def hello_world_pipeline():
    task = example_comp(
        name="hello world!"
    )

Elyra

See: https://elyra.readthedocs.io/en/latest/user_guide/pipeline-components.html#managing-pipeline-components

Creating Components

Each subfolder of this repository includes a Kubeflow component. Each component is generally structured as recommended by Kubeflow's "Building Components" guide. That is, each component comes with:

a component.yaml file that specifies the component’s interface and implementation and
(optionally) a build_image.sh file that can be run to build a component-specific container image and to push it to a container registry.

Base Images

The base-images folder includes only build scripts for images that some components might depend on. There's no component.yaml for base images themselves.

Building Component-Specific Container Images

Some components may require component-specific container images, which have to be build first. In this case, proceed as follows.

git clone https://github.com/lehrig/kubeflow-ppc64le-components
cd kubeflow-ppc64le-components

Afterwards, cd into the component in question and run:

docker login ***YOUR TARGETED IMAGE REGISTRY***
./build_image.sh