Introduction

IMPORTANT NOTICE: As of April 8th, 2021, we are migrating and reworking our docs for a better developer experience. We expect the migration to be complete by April 15th.

During the migration, you might experience broken links, images and more. We apologize for the inconvenience. Please bear with us as we make the docs a better developer experiece for you. Thank you!

ZenML is an extensible, open-source MLOps framework for using production-ready Machine Learning pipelines, in a simple way. At its core, ZenML will orchestrate your experiment pipelines from sourcing data to splitting, preprocessing, training, all the way to the evaluation of results and even serving.

While there are other pipelining solutions for Machine Learning experiments, ZenML is focussed on two unique approaches:

Why do I need ZenML?

ZenML solves the problem of getting Machine Learning in models. You should use ZenML if you struggle with:

  • Reproducing training results in production.

  • Managing ML metadata, including data, code, and model versioning.

  • Getting (and keeping) ML models in production.

  • Reusing code/data and reducing waste.

  • Maintaining comparability between ML models.

  • Scaling ML training/inference to large datasets.

  • Retaining code quality alongside development velocity.

  • Keeping up with the ML tooling landscape in a coherent manner.

Awesome things you can do with ZenML

  • Reproduce experiments at any time, on any environment. [here's how].

  • Automatically track all parameters when creating ML experiments. [here's how].

  • Collaborate with your team using a git repo, re-use code, share results and compare experiments. [here's how].

  • No-hassle preprocessing and training on big VM's on the, for 1/4th the price. [here's how].

  • Distribute preprocessing on hundreds of workers for millions of datapoints. [here's how].

  • Launching training jobs on GPUs on the cloud with a simple command. [here's how].

  • No-hassle evaluation of models with slicing metrics. [here's how].

  • Instantly deploy a model to the cloud. [here's how].

  • De-couple infrastructure from ML code. [here's how].

What do I do next?

If one of the above links are too hands-on, then a good place to go from this point is:

Get involved

If you're just not ready to use ZenML for whatever reason, but still would like to stay updated, then the best way is to star the GitHub repository! You can then keep up with the latest going-on's of ZenML, and it would help us tremendously to get more people using it.

Contributions are also welcome! Please read out contributing guide to get started.