Create E2E example of distributed model training and serving.

[elsonrodriguez]

Currently most of the examples do not show how to complete the training of a model within kubeflow, and also take that trained model and serve it with kubeflow.

We need an example that covers this from start to finish.

[elsonrodriguez]

Working on this, can someone assign?

[jlewi]

This is fantastic. What dataset are you working with? What is the ML task?

[jlewi]

@elsonrodriguez Ping?

[elsonrodriguez]

We're trying to narrow down what model/data combo can be fully trained within a reasonable timeframe.

Any suggestions?

[texasmichelle]

@elsonrodriguez I'm beginning work on Kubeflow #157. Want to help? I broke it down into pieces as issues in this repo, so feel free to pick one of them up.

[elsonrodriguez]

@texasmichelle Yeah Sure! We ended up going with mnist and are just about done with an initial pass on this. I'll ping you when I open the PR and if you see anything that can be re-used let me know and I'll try to make it more generic to fit the GH summarization use case.