Add examples for `sdk` for users to try

[jaiakash]

What you would like to be added?

This issue tracks the addition of few user-facing examples to the Kubeflow SDK repository, similar to the examples already available in the trainer repository.

Kubeflow Trainer examples :

• https://github.com/kubeflow/trainer/tree/master/examples

Why is this needed?

• Helping new users quickly understand how to use the SDK with minimal setup
• Providing ready-to-run examples for experimentation and learning
• Establishing standard patterns and best practices for SDK usage

Love this feature?

Give it a 👍 We prioritize the features with most 👍

[jaiakash]

/kind discussion
/good-first-issue
/assign

[google-oss-prow]

@jaiakash:
This request has been marked as suitable for new contributors.

Please ensure the request meets the requirements listed here.

If this request no longer meets these requirements, the label can be removed
by commenting with the /remove-good-first-issue command.

Details

In response to this:

/kind discussion
/good-first-issue
/assign

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[ananyajain10]

Hi @jaiakash ,
I’m interested in contributing to this issue by adding beginner-friendly SDK examples.
Could you please assign this issue to me?

Thank you.

[jaiakash]

Hi @ananyajain10 can you suggest what and which examples you would like to add here?
For reference, you can check out Kubeflow Trainer examples :

kubeflow/trainer@master/examples

For now, I am NOT assigning you, as this issue is still under discussion, but I would love to see what examples you propose.

[ananyajain10]

Hi @jaiakash

I would like to propose a deep learning NLP example to demonstrate Kubeflow’s usefulness for transformer and LLM-style workloads.

Proposed example:
“BERT Sentiment Analysis with Kubeflow SDK”

Problem:
Fine-tune a pretrained BERT/DistilBERT model to classify IMDB movie reviews as positive or negative.

What this example would demonstrate:
Fine-tuning a Hugging Face transformer model using PyTorch
GPU-based training with the Kubeflow SDK
Submitting and managing a TrainJob via the SDK
Monitoring logs and job status, and saving the trained model

With reference to the provided examples, this is a real-world NLP use case with a moderate model size, making it suitable for beginners to get started with Kubeflow.
If this aligns with the direction of the issue, I’d be happy to proceed with implementing this example.

Thank you.

[jaiakash]

Sure @ananyajain10 go ahead.

/remove-label lifecycle/needs-triage
/assign @ananyajain10

[andreyvelich]

Thank you for creating this @jaiakash and @ananyajain10 for taking a look.

I would suggest first taking a look at this:

• chore: Add Speech Recognition with DDP Example trainer#2830
• Add more AI/ML Training Example with Kubeflow Trainer trainer#2040

@ananyajain10 As a first step you can take over the 2830 PR to finalize the speech recognition example.

Since we don't have E2E infra in kubeflow/sdk yet, I would suggest that we add example to kubeflow/trainer for now.