⚠ This page documents a version of Oakestra which is not the latest stable. Please refer to the latest docs for a current version.

Example Projects

Check out these examples from the community!

Here a collection of examples that you can already use to experiment with FLOps.

Handwritten Digits

For the handwritten digits recognition task, we provide 4 different examples based on the mnist training dataset. For the training, you can use the default mock data provider.

  curl -sSl https://oakestra.io/FLOps_SLAs/mocks/mnist_multi.json > mnist_multi.json
  oak addon flops mock-data mnist_multi.json

Do you want to injest your own data?

FLOps uses Arrow Flight for data ingestion. Each Learner Node uses a Arrow Flight server and client combination. Check out here how this works.

This is a small example also used as the example in the [../../flops-project-workflow/overview] section.

You can download the SLA for this example using

  curl -sSl https://oakestra.io/FLOps_SLAs/projects/mnist_sklearn_small.json > mnist_sklearn_small.json

and then you can run the project using

  oak addon flops p mnist_sklearn_small.json

You can download the SLA for this example using

  curl -sSl https://oakestra.io/FLOps_SLAs/projects/mnist_sklearn_large.json > mnist_sklearn_large.json

and then you can run the project using

  oak addon flops p mnist_sklearn_large.json

You can download the SLA for this example using

  curl -sSl https://oakestra.io/FLOps_SLAs/projects/hierarchical_mnist_sklearn_small.json> hierarchical_mnist_sklearn_small.json

and then you can run the project using

  oak addon flops p hierarchical_mnist_sklearn_small.json

You can download the SLA for this example using

  curl -sSl https://oakestra.io/FLOps_SLAs/projects/hierarchical_mnist_sklearn_large.json > hierarchical_mnist_sklearn_large.json

and then you can run the project using

  oak addon flops p hierarchical_mnist_sklearn_small.json

Core differences between these examples

Namemoderoundsclientscyclespost_trainingarch
Smallclassic32N/Abuild & deploy trained modelamd64
Largeclassic104N/Abuild & deploy trained modelamd64
Hierarchical Shierarchical322build & deploy trained modelamd64
Hierarchical Lhierarchical5310build & deploy trained modelamd64

Check out FLOps SLA configuration page to learn how to customize your project.

Object Recognition

For a simple FL object recognition example, here you can find an example based on the cifar10 dataset.

For the training, you can use the default mock data provider for this example.

  curl -sSl https://oakestra.io/FLOps_SLAs/mocks/cifar10_simple.json > cifar10_simple.json
  oak addon flops mock-data cifar10_simple.json

Do you want to injest your own data?

FLOps uses Arrow Flight for data ingestion. Each Learner Node uses a Arrow Flight server and client combination. Check out here how this works.

Then, you can download the SLA for this example running

  curl -sSl https://oakestra.io/FLOps_SLAs/projects/cifar10_pytorch.json > cifar10_pytorch.json

and finally you can run the project using

  oak addon flops p cifar10_pytorch.json

Here a summary of the settings for this default SLA

Standard Example Configuration

Namemodeflavorroundsclientspost_trainingarch
cifar10_pytorchclassicpyTorch32build & deploy trained modelamd64

Check out FLOps SLA configuration page to learn how to customize your project.

Prev
Development
Next
Features