[ANNOUNCE] Kubeflow SDK 0.5 is released!

[Andrey Velichkevich]

Hi Folks,

I'm thrilled to announce the release of Kubeflow SDK 0.5.0 🎉

pip install kubeflow==0.5.0

Highlights

✅ New PipelinesClient() to create and manage Kubeflow Pipelines from the unified Kubeflow SDK with the simplified and consistent APIs

✅ SparkClient now supports Spark batch jobs – submit a Python function directly with FuncJob, or run an existing Spark application file from local or remote storage with FileJob

✅ Trainer: TPU multi-slice support with automatic numNodes calculation

✅ Kubeflow MCP Server page: https://sdk.kubeflow.org/en/latest/mcp-server/index.html

Learn more about the recent updates in these resources

1. Kubeflow SDK Evolution – One Million Downloads and Counting

2. Batch Jobs for SparkClient – Submitting and Managing Spark Workloads from Python

3. The Kubeflow SDK 0.5 release notes

If you have any questions, please reach out in the #kubeflow-ml-experience Slack channel or join the Kubeflow SDK bi-weekly calls.

Regards,

Andrey