kfp api #10

Supports: kubernetes

Deploy this charm on Kubernetes with the CLI. Find out how by reading the docs.

Description

Machine learning (ML) toolkit that is dedicated to making deployments of ML workflows on Kubernetes simple, portable, and scalable.


Kubeflow Pipelines API Operator

Overview

This charm encompasses the Kubernetes Python operator for Kubeflow Pipelines API (see CharmHub).

Install

To install Kubeflow Pipelines API, run:

juju deploy kfp-api

For more information, see https://juju.is/docs


Configuration

auto-update-default-version
(boolean) If true, the default pipeline version will be updated when uploading a new version of a pipeline
True
cache-enabled
(boolean) If true, pipeline run steps can be cached instead of re-run
True
cache-image
(string) Which image to list as the backing image for a pipeline run step pulled from cache
gcr.io/google-containers/busybox
grpc-port
(string) GRPC port
8887
http-port
(string) HTTP port
8888
init-connection-timeout
(string) Connection timeout used when initializing clients for related services. The format used can be anything accepted by `time.ParseDuration`.
5s
log-archive-filename
(string) Name of log file in object storage
main.log
log-archive-prefix
(string) Prefix for log file in object storage
/artifacts
runner-sa
(string) Default pipeline runner service account. Used if service account is left unspecified when creating a run
pipeline-runner