Examples

The following examples demonstrate how to use the Kubeflow SDK for distributed AI training and LLM fine-tuning.

PyTorch & HuggingFace Examples

Task	Model	Dataset	Notebook
Local Training	CNN	MNIST	Open Notebook
Image Classification	CNN	Fashion MNIST	Open Notebook
Question Answering	DistilBERT	SQuAD	Open Notebook
Speech Recognition	Transformer	Speech Commands	Open Notebook
Audio Classification	CNN (M5)	GTZAN	Open Notebook

DeepSpeed Examples

Task	Model	Dataset	Notebook
Text Summarization	T5	CNN/DailyMail	Open Notebook

MLX Examples

Task	Model	Dataset	Notebook
Image Classification	MLP	MNIST	Open Notebook
LLM Fine-Tuning	Llama 3.2-3B	WikiSQL	Open Notebook

TorchTune Examples

Task	Model	Dataset	Notebook
LLM Fine-Tuning	Llama 3.2-1B	Alpaca	Open Notebook

Spark Examples

Task	Example	Description	Notebook
Basic Spark Client	SparkClient API	Basic SparkClient usage with a simple API	Open Notebook
Advanced Spark Configuration	Driver / Executor Configuration	Configure Spark jobs using Driver and Executor objects	Open Notebook
Connect to Existing Spark Cluster	Existing Spark Connect Session	Connect to an already running Spark Connect cluster	Open Notebook
Interactive Spark Connect	Live SparkConnect Demo	Provision, inspect, and manage SparkConnect sessions	Open Notebook
Batch Job Lifecycle	Spark Batch Job	Submit, monitor, inspect logs, and delete Spark batch jobs	Open Notebook
Python Function Job Lifecycle	Spark FuncJob	Run standalone Python functions as distributed Spark jobs	Open Notebook
Batch Job Options	Advanced Batch Configuration	Configure batch jobs with labels, annotations, and node selectors	Open Notebook
Failed Batch Job Handling	Error Diagnostics & Logs	Inspect failure statuses and driver pod logs for failed batch jobs	Open Notebook