Cake provides a variety of Kubeflow Pipeline (KFP) Components and pure-python Components suitable for workflow engines like Airflow. These Components help you ML workflows integrating a variety of functionality.
Installation
Pipeline Components are available to customers via a customer-specific Git Repository. Setting this up properly requires a little help from one of our Solution Engineers. Please reach out to set up a 30 minute meeting to get going.
Overview of Available Components
Component Name | Description |
|---|---|
artifact_to_shared | Exports a Kubeflow Pipeline artifact file (or directory) to destination on the Cake shared drive |
autoviz_component | Runs AutoViz given the specified input_data and dependent variable for analysis |
get_namespace_component | Returns the Kubernetes Namespace |
kserve_create_endpoint | Creates a KServe InferenceService Endpoint |
mlflow_download_artifacts | Downloads Model Artifacts from MLflow |
send_slack_message | Sends a message to a slack channel |
shared_to_artifact | Imports data from /home/jovyan/shared into a Kubeflow Pipeline artifact |
Example Kubeflow Pipelines
We provide example KFP Pipelines that showcase how to use these components:
Pipeline Name | Description |
|---|---|
autoviz_pipeline | Creates a simple Pandas dataframe and runs the Autoviz Component |
get_namespace_pipeline | Gets the Kubernetes Namespace |
mlflow_to_kserve | Downloads an MLFlow Model using the MLflow Component then creates a KServe Endpoint |
send_slack_message_pipeline | Sends a message to an authorized slack channel |